Research ArticleExploring Project Complexity through Project Failure FactorsAnalysis of Cluster Patterns Using Self-Organizing Maps

Vicente Rodriacuteguez Montequiacuten Joaquiacuten Villanueva BalseraSonia Mariacutea Cousillas Fernaacutendez and Francisco Ortega Fernaacutendez

Project Engineering Area University of Oviedo CIndependencia 13 33004 Oviedo Spain

Correspondence should be addressed to Vicente Rodrıguez Montequın montequiuniovies

Received 27 October 2017 Accepted 2 April 2018 Published 22 May 2018

Academic Editor Luis Carral Couce

Copyright copy 2018 Vicente Rodrıguez Montequın et al This is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

In the field of project management complexity is closely related to project outcomes and hence project success and failurefactors Subjectivity is inherent to these concepts which are also influenced by sectorial cultural and geographical differencesWhile theoretical frameworks to identify organizational complexity factors do exist a thorough and multidimensional accountof organizational complexity must take into account the behavior and interrelatedness of these factors Our study is focused onanalyzing the combinations of failure factors by means of self-organizing maps (SOM) and clustering techniques thus gettingdifferent patterns about the project managers perception on influencing project failure causes and hence project complexity Theanalysis is based on a survey conducted among project manager practitioners from all over the world to gather information on thedegree of influence of different factors on the projects failure causes The study is cross-sectorial Behavioral patterns were foundconcluding that in the sampled population there are five clearly differentiated groups (clusters) and at least three clear patternsof answers The prevalent order of influence is project factors organization related factors project manager and team membersfactors and external factors

1 Introduction

As projects have become more and more complex there isan increasing concern about the concept of project com-plexity and its influence upon the project management pro-cess Projects have certain critical characteristics that deter-mine the appropriate actions to manage them successfullyProject complexity (organizational technological informa-tional etc) is one such project dimensionTheproject dimen-sion of complexity is widespread within project managementliterature In the field of projects complexity is closed relatedto the causal factors to get the project outcomes which inthe project management field is usually referred to as ldquoprojectsuccessfailure factorsrdquo Better understanding of project suc-cessfailure factors is a key point for creating a strategy tomanage complexity

Most attributes of complexity are known to be constantlychanging variables such as project type project size projectlocation project team experience interfaces within a project

logisticsmarket conditions geopolitical and social issuesand permitting and approvals [1] Several studies focus onproject complexity and the factors that influence its effecton project success In general there is not a clear differencebetween complexity and success factors when consideringthe literature of project management For instance Gidadodefined project complexity and identified the factors thatinfluence its effect on project success in relation to estimatedproduction time and cost based on literature search andstructured interviewing of practitioners [2] Kermanshachiet al [3] consider that when complexity is poorly under-stood and managed project failure becomes the norm Theyfocused on strategies to manage complexity in order toincrease the likelihood of project success Vidal andMarle [4]define project complexity as a property of a project whichmakes it difficult to understand foresee and control theprojectrsquos overall behavior Remington et al [5] believe that acomplex project demonstrates a number of characteristics toa degree or level of severity that makes it difficult to predict

HindawiComplexityVolume 2018 Article ID 9496731 17 pageshttpsdoiorg10115520189496731

2 Complexity

project outcomes or manage projects One of the projectcomplexity definitions that best fits the aim of this study wasthe one given by Kermanshachi et al ldquoProject complexity isthe degree of interrelatedness between project attributes andinterfaces and their consequential impact on predictabilityand functionalityrdquo [3] This definition can guide the iden-tification of project complexity indicators and managementstrategies which reduce the undesired outcomes often relatedto project complexity Under this context we can expecta relationship among project failure causes (derived fromproject attributes and interfaces) and project complexity

Success in projects is more complex than just meetingcost deadlines and specifications In fact customer satisfac-tion with the final result has a lot to do with the perception ofsuccess or failure in a project In the end what really mattersis whether the parties associated with and affected by a projectare satisfied or not [6] Meeting deadlines and costs doesnot really matter if the final project outcomes do not meetexpectations

However our work is not focused on the concepts ofsuccess or failure but on the study of the aspects that leadto the failure in projects and their combinations consideringthat the complexity of a project is determined by these factorsThere are plenty of factors whose application is significantfor the success or failure of a project In the literature theseare called critical success factors and many studies have beendevoted to define clarify and analyze such factors Successfactors are subjected to the perceptions of the ones involvedin the project development depending not only on the stake-holder but also on cultural or geographical differences whichare reflected in the context of the organization [7] There arealso a lot of sectorial influences For example Huysegoms etal have identified the causes of failure in software productlines context [8] Other examples can be found in [9ndash12]

Obviously projects fail due to many different reasonsif we understand ldquofailurerdquo as the systematic and widespreadnoncompliance of the criteria which define a successfulproject [13] Nevertheless due to the inner subjectivity of theconcept each person working in the same project has apersonal opinion about the determining causes of its failureThese opinions can also vary depending on the type andsector of projects so that distinctive patterns of causes areassociated with the failure of specific kinds of projects Themost usual is that a combination of several factors withdifferent levels of influence in different stages of the life cycleof a project result in its success or failure

Most of the studies in the literature focus on determininglists of factors and their categorization ranking them accord-ing to their influence level Nevertheless projects behaviorderives from ldquosystematic interrelated sets of factorsrdquo ratherthan single causal factors and the fact that true causes ofproject outcomes are difficult to identify [14 15] The interac-tions between the different factors seem to be as important aseach factor separately However there seems to be no formalway to account for these interrelations which may be thereason why this point is weakly treated in the literature Ourstudy is focused on analyzing the combination of these factorsby means of clustering techniques thus getting differentpatterns about the project manager practitioners perception

on influencing project failure causes and hence projectcomplexity

The work here presented is part of a more global studyanalyzing success factors and failure causes in projects Aquestionnaire has been aimed specifically to project man-agement practitioners to gather information on the degreeof influence of different factors on the failure or success ina project The questionnaire inquires about their perceptionon the most influential factors to be considered to reachsuccess as well as the common failure causes they havemost frequently encountered A selection of critical successfactors and failure causes were selected as a basis for thequestionnaire compiling previous research work results [1617] with the most frequent causes reflected in the literatureThe questionnaire is generic not intended for any specificsector or geographical area Although it is not focusedon any particular project or field it gathers this type ofinformation to be able to correlate it The survey was dis-tributed anonymously to recipients through LinkedIn anInternet professional network Such study determines themost frequent failure causes and the most important successfactors in real world projects In the initial stage a statisticalanalysis of the sample data was conducted with the aim ofanswering the question of whether the valuations dependon the geographical areas of the respondents or on thetypes of projects that have been carried out [16] The studyshows that there is no absolute criterion and that subjectivityis the inherent characteristic of those valuations So as acomplementary study clustering techniques are applied inorder to find patterns in the set of received answers Thispaper presents the obtained results Our study is a cross-sectorial study which is a remarkable improvement over themajority of studies that consider only one sector with theconstruction one being of the most studied

The remainder of this paper is organized as followsSection 2 reviews the literature related to the environmentof this work Section 3 describes the research methodologySections 4 and 5 present and discuss the results and finallySection 6 exposes the conclusions

2 Literature Review

The literature is explored according to the three main pointsof this work background of complexity in the field ofproject management project successfailure causes and theirconnection with project complexity and applications of self-organizing maps (SOM) in the field of project management

21 Project Complexity Theory Complexity has been recog-nized as one of the most relevant topics in project manage-ment research According to Baccarini [18] one of the firstreviews about complexity in the project management fieldproject complexity can be defined in terms of differentiationand interdependency and it is managed by integration Inthe definition differentiation refers to the number of variedcomponents of the project (tasks specialists subsystemsand parts) and interdependency refers to the degree ofinterlinkages among these components He established thedichotomy considering that project complexity is composed

Complexity 3

of technological complexity and organizational complexityComplexity can take various forms namely social technolog-ical environmental and organizational Worth mentioninghere is the work of Bosch-Rekveldt et al [19] who proposedthe TOE framework consisting of fifty factors in threefamilies technical organizational and environmental Theauthors have also concluded that organizational complexityworried project managers more than technical or environ-mental complexities Vidal andMarle [4] argued that approx-imately 70 of project complexity factors are organizationalWe follow these assertions by focusing our study mainly onthe organizational factors knowing that this covers a majorarea of complexity in projects

Scholars have focused on the identification of complexityattributes more than any other topic in the field of projectcomplexity Studies in this area have evolved significantly overthe past twenty years Cicmil et al [20] identified complexityas a factor that helps determine planning and control prac-tices and hinders the identification of goals and objectives ora factor that influences time cost and quality of a projectGidado [2] defined project complexity and identified thefactors that influence its effect on project success Also thestudy proposes an approach that measures the complexity ofthe production process in construction

Project complexity has been analyzed in the book editedby Cooke-Davies [21] from three different perspectives peo-ple who manage programs and projects (practitioners) linemanagers in organizations to which programs and projectsmake a substantial contribution (managers) andmembers ofthe academic research community who have an interest inhow complexity shapes and influences the practice of pro-gram and project management (researchers) The book con-stitutes a valuable resource to put together what is currentlyknown and understood about the topic to help practitionersand their managers improve future practice and to guideresearch into answering those questions that will best help toimprove understanding of the topic

Althoughmost authors emphasized the influence of inter-dependencies and interactions of various elements on projectcomplexity [22] few works analyze those dependencies Forexample TOE model [19] does not allow for an understand-ing of how various elements contribute to overall complexityNevertheless other authors regarded project complexity ashaving nonlinear highly dynamic and emerging featuresVidal et al [23] for example proposed the definition ofproject complexity as ldquothe property of a project which makesit difficult to understand foresee and keep under controlits overall behavior even when given reasonably completeinformation about the project systemrdquo Lu et al [24] proposethat project complexity can be defined as ldquoconsisting of manyvaried interrelated parts and has dynamic and emergingfeaturesrdquo Lessard et al built the House of Project Complexity(HoPC) a combined structural and process-based theoreticalframework for understanding contributors to complexity

The connection between failure and complexity in projectmanagement has also been established in the literature Ivoryand Alderman [22] studied the failure in complex systemsin order to shed some critical light on the management ofcomplex projects The authors conclude that it was not the

technical complexity per se that made the management ofthe projects complex Rather the primary determinants ofthe complexity of the project management process stemmedfrom changes in the markets regulatory context and knowl-edge requirements facing the project

There are connections between the project complexityindicators identified in the literature and the project failurecauses For example Kermanshachi et al [3] identified 37indicators Among them several examples usually foundin the literature as project failure causes are included (ieimpact of external agencies on the project execution planimpact of required approvals from external stakeholderslevel of difficulty in obtaining permits level of project designchanges derived by Request for Information (RFI) etc) Infact there are several examples of authors that are referencedfrequently in the field of both project complexity and projectsuccess factors Shenhar et al [25] Dvir and Lechler [26]Cooke-Davies [27] and Pinto and Prescott [28] are someexamples of those

22 Project SuccessFailure Factors One of the most relevantfields of study in project management is success factors andfailure causes in projectsTheywere first identified at the PMIAnnual Seminar and Symposium in 1986 [29] and becameone of themost discussed themeswithin specialized literature[30] Each stakeholder working on a project has hisher ownopinion on what is determinant for failure and it is muchmore complex than adhering to the traditional criteria oftime cost and quality If a project can be considered failedwhen it has not delivered what was expected after its com-pletion causes leading to this unsatisfactory result are alsosubjected to the different points of view of the stakeholdersinvolved

The different lists of success factors and failure causesin the literature show no consensus The most usual wouldbe a combination of several factors with different levels ofinfluence in different stages of the projectrsquos life cycle resultingin its success or failure The concepts of success factors andfailure causes are closely related but a failure cause is notnecessarily the negation of or the opposite to a successfactorThere is not always such correspondence among themThis study is considering failure causes rather than successfulfactors because we are taking the assumption that the morethe failure causes concur in a project the higher the complex-ity is

The literature identifying successfailure factors is veryextensive Some examples especially regarding failure causesare [31ndash35] all of them related to the construction sectorAnother sector with many references is the IT (informationtechnology) projects Some examples are [36ndash39] There areseveral frameworks classifying the factors Belassi and Tukel[40] suggested a scheme that classifies the critical factorsin four different dimensions (Table 1) and describes theimpacts of these factors on project performance Shenhar etal [25] have identified also four distinct dimensions projectefficiency impact on the customer direct and business suc-cess and preparing for the future They stated that the exactcontent of each dimension and its relative importancemay change with time and are contingent on the specific

4 Complexity

Table 1 Belassi and Tukel taxonomy

Groups of factorsRelated to projectRelated to the project managers and the team membersFactors related to the organizationFactors related to the external environment

stakeholder Lim and Mohamed [41] consider two view-points of project success macro and micro Regarding themicro they have identified technical commercial financerisk environmental and human related factors

Richardson [42] and King [43] point out that none of thekey success factors described in the literature is responsibleon its own for ensuring a projectrsquos success They are allinterdependent and require a holistic approach Groups ofsuccess factors and their interactions are of prime importancein determining a projectrsquos success or failure [44] Multivariatestatisticsmethodsmay be very useful with this purpose Someexamples applied in the project management complexity are[4 45 46] as well as in the field of project success [47ndash49]Clustering methods have been also used in the context ofexploring complex relationships within the field of projectmanagement for example the works [50ndash52] Our study usesself-organizing maps (SOM) and clustering techniques tofind patterns in a data set of answers coming from a survey

23 Self-Organizing Maps and Applications in Project Man-agement SOM is an unsupervised neural network proposedby Kohonen [53] for visual cluster analysis The neurons ofthe map are located on a regular grid embedded in a low(usually 2 or 3) dimensional space and associated with thecluster prototypes by the connected weights In the course ofthe learning process the neurons compete with each otherthrough the best-matching principle in such a way that theinput is projected to the nearest neuron given a defineddistance metric The winner neuron and its neighbors on themap are then adjusted towards the input in proportion withthe neighborhood distance consequently the neighboringneurons likely represent similar patterns of the input dataspace Due to the data clustering and spatialization throughthe topology preserving projection SOM is widely usedin the context of visual clustering applications Despite theunsupervised nature the applicability of SOM is extendedto classification tasks by means of a variety of ways suchas neuron labeling method semisupervised learning orsupervised learning vector quantization (LVQ) [54]

SOM is recognized as a useful technique to analyze high-dimensional data sets and understand their hidden relationsIt can be used to manage complexity in large data sets [39]Nevertheless there are few cases described in the field ofproject management Balsera et al [55] have exposed theapplication of SOM to analyze information related to effortestimation and software projects features MacDonell [56]has reported other multidimensional data study visualizationbased on SOM identifying groups of data for similar projectsand finding nonlinear relationships within the exploredvariables Naiem et al [57] have used SOM for visualizing the

Table 2 Failure causes considered in the study

Cause CodeIrruption of competitors C1Continuous or dramatic changes to the initial requirements C2Customerrsquos requirements inaccurate incomplete or notdefined C3

Disagreement or conflict of interest among departments C4Inaccurate cost estimations C5Inaccurate time estimations C6Deficient management of suppliers and procurement C7Lack of Management support C8Lack of previous identification of legislation C9Badly defined specifications C10Political social economic or legal changes C11Project manager Lack of commitment C12Project manager Lack of communication skills C13Project manager Lack of competence C14Project manager Lack of vision C15Project requirements deficiently documented C16Project staff changes C17Project team lack of competence C18Project team misunderstanding related to customeruserneeds C19

Project team lack of commitment C20Public opinion opposition to Project C21Quality checks badly performed or not performed at all C22Extremely new or complex technology C23Unexpected events with no effective response possible C24Unrealistic customer expectations C25Wrong number of people assigned to the project C26

set of candidate portfolios After reviewing the literature wecan conclude that there is not any former application of SOMfor analyzing patterns of project failure causes

3 Research Methodology

As discussed in the Introduction the ground for this workis the questionnaire that was designed to gather informationon the perception project managers have of what the successfactors and failure causes are After the information wasgathered a multivariate analysis was performed on the datawith cluster data mining techniques

The sections of the questionnaire considered for thisstudy were as follows

(i) General information on the respondent and typologyof projects heshe was involved in country type andsize of project

(ii) Frequency of different causes of project failure with26 multiple-choice questions (from 0ndash25 rare orimprobable occurrence to 75ndash100 always occurs)

The causes of failure are extracted from the existingbibliography and the previous work on the matter Table 2

Complexity 5

Factors related to organization

middotDisagreement or conflict of interest among departments (C4)

middotDeficient management of suppliers and procurement (C7)

middotLack of Management support (C8)

middotProject staff changes (C17)

middotWrong number of people assigned to the project (C26)

Factors related to project

middotContinuous or dramatic changes to the initial requirements (C2)middotCustomerrsquos requirements inaccurate incomplete or not defined (C3)middotInaccurate cost estimations (C5)middotInaccurate time estimations (C6)middotLack of previous identification of legislation (C9)-Badly defined specifications (C10)middotProject requirements deficiently documented (C16)middotQuality checks badly performed or not performed at all (C22)middotExtremely new or complex technology (C23)

Factors related to the project manager and team members

middotProject manager Lack of commitment (C12)

middotProject manager Lack of communication skills (C13)

middotProject manager Lack of competence (C14)

middotProject manager Lack of vision (C15)

middotProject team lack of competence (C18)

middotProject team misunderstanding related to customeruser needs (C19)

middotProject team lack of commitment (C20)

Factors related to external environment

middotIrruption of competitors (C1)

middotPolitical social economic or legal changes (C11)

middotPublic opinion opposition to Project (C21)

middotUnexpected events with no effective response possible (C24)

middotUnrealistic customer expectations (C25)

FAILURE

Figure 1 Failure causes grouped by category The factors are derived from existing bibliography and previous work on the matter They havebeen classified according to Belassi and Tukel taxonomy

gathers the identified factors The factors are coded as 119862119909where 119909 is the number given to the failure cause

The factors were classified according to Belassi andTukel taxonomy [40] included in Table 1 Figure 1 depictsthe classification considered The factors are presented un-shorted because they were shuffled in order to avoid biasesin the survey The first group includes the factors related tothe organization the project belongs to (ie factors relatedto top management support) The second group includes thefactors related to the project itself and the way the projectis managed The third one comprises the factors related tothe project manager and the team members Many studiesdemonstrated the importance of selecting project managerswho possess the necessary technical and administrativeskills for successful project termination as well as theircommitmentThe competences of the teammembers are alsofound to be a critical factor Finally the factors related toexternal environment consist of factors which are external tothe organization but still have an impact on project successor failure The classification can be considered collectivelyexhaustive Belassi and Tukel state that the four groupsoffer a comprehensive set in that any factor listed in the

literature or even specific points of consideration shouldbelong to at least one group [40] However it is not easyto differentiate in which category to include each of thedifferent factors The groups are interrelated The authorsgive some indications to distinguish where to classify themmainly regarding organization and external environmentFor example if a customer is from outside the organizationheshe should be considered as an external factor (ie C25-unrealistic customer expectations) For functional projectshowever customers are usually part of the organization suchas topmanagement In such cases factors related to the clientcan be grouped under the factors related to the organization

We can find correspondences among most of the failurecauses included in our study (Table 2) and the complexityindicators appointed by Kermanshachi et al [3] For exampleC2 (which is one of the most frequent in accordance withthe rank presented in Table 7) is related to several complexityindexes as ldquolevel of project design changes derived by RFIrdquoand ldquomagnitude of change ordersrdquo C10 for example isrelated to ldquopercentage of design completed at the start ofconstructionrdquo This fact reinforces the connection betweenfailure causes and complexity factors

6 Complexity

The recipients were randomly chosen among the mem-bers of 36 project management groups from LinkedIn net-work The questionnaire was open for 3 months and 11 daysin 2011 in order to obtain a significant number of answersDuring that period of time customized emails were sentwith an invitation to answer the questionnaire to a total of3668 people A total of 619 answers were received (1688)611 of which were considered for further analysis (the restwere discarded due to consistency issues) Neither of thequestionnaires was partially filled or incomplete since allfields were marked as mandatory Previously in 2010 a pilotsurvey was conducted with the help of project managementexperts and practitioners This primary questionnaire wassent up to 45 people with a response rate of 6667 Thosefactors which reached a higher score were the ones finallyselected to configure the list included in the current studySuggestions and comments made by respondents to improvethe proposed list were also taken into account

Answers from 63 countries were received as Table 3shows

The answers have been grouped first into 13 differentgeographical zones (more details can be found in [16])according to geographical cultural historical and economiccriteria From those a total of 6 groups comprise more than85 of the respondents AMZ3 EUZ3 EUZ1 AMZ1 ASZ1and ASZ3 The other 15 of the responses were grouped in anew category called ldquoOthersrdquo Table 4 summarizes the groupsby geographical zone

Following a similar approach 53 types of projects werepresent on the answers received To simplify they have beenclassified into a total of 17 groups according to the highestlevel of the ISIC Rev 4 Classification [58] Development aidprojects were listed apart as an independent category Table 5presents the number of respondents from each project typeand the codification

Information about the size of the projects the respondentsare usually involved in was also enquired The size of theprojects is something quite subjective anddifficult to compareamong different sectors For example considering the projectbudget as an indicator of project size a project consideredsmall in construction could by contrast be consideredlarge in information technology projects In order to avoidthese biases the guideline provided by the University ofWisconsin-Madison [59] was used The results show that1195 of respondents are involved in small projects 4926in medium size projects and 3879 in large projects

Though the scope of this paper does not cover thestatistical analysis of the obtained answers here is a summaryof the main results Overall both the most and least frequentfailure causes are presented here In order to rank each one ofthe 26 failure causes a frequency index (FI) was calculated asfollows

FI = [sum611119894=1 sum4119895=1 119860 119894 lowast 119861119895]611 (1)

where119860 119894 is the number of responses choosing interval 119894119861119895 is interval factor 119895

Table 3 Countries and number of responses (ordered by numberof responses) When several countries are indicated in the samerow the number of respondents stands for the number of answersreceived from each country (ie 50 answers from Spain and 50answers from the United States)

COUNTRIES NUMBER OFRESPONDENTS

Argentina 91Spain United States 50Greece 45Chile 41India 40Brazil 30Luxembourg 27Mexico 21Uruguay 20United Arab Emirates 19Italy 18Russia 16Sweden 14South Africa 12Canada 11Germany 8Saudi Arabia The Netherlands UnitedKingdom 7

Australia Denmark Finland Kuwait 5Belgium Paraguay Slovenia 4Egypt France Israel Switzerland 3Philippines Tanzania 2Algeria Austria Bahrain Belarus BotswanaBritish Virgin Islands Cameroon ChinaColombia Cyprus Czech Republic GhanaHong Kong Hungary Indonesia IrelandJapan Jordan Nigeria Pakistan Peru PolandQatar Rwanda Singapore Surinam ThailandUganda Venezuela

1

Each interval factor is defined as presented in Table 6 Theranking is presented in Table 7

With the obtained results a cluster analysis was con-ducted to obtain patterns in the answers data set groupinga set of data with similar values The aim is to classify a set ofsimple elements into a number of groups in such a way thatthe elements in the same group are similar or related to oneanother and at the same time different or unrelated to theelements in other groups

The cluster analysis (also called Unsupervised Classi-fication Exploratory Data Analysis Clustering NumericalTaxonomy or PatternRecognition) is amultivariate statisticaltechnique whose aim is to divide a set of objects into groupsor clusters in such a way that objects in the same cluster arevery similar to each other (internal cluster cohesion) andthe objects in other groups are different (external clusterisolation) Summarizing it deals with creating data clusters

Complexity 7

Table 4 Answers grouped by geographical zone

GEOGRAPHICAL ZONESANALYZED COUNTRIES NUMBER OF

RESPONDENTSAMZ3 Argentina Brazil Chile Colombia Paraguay Peru Surinam Uruguay and Venezuela 190EUZ3 Cyprus Spain Greece and Italy 114

EUZ1 Germany Austria Belgium Denmark Finland France Ireland Virgin Islands (UK)Luxembourg United Kingdom The Netherlands Sweden and Switzerland 86

AMZ1 Canada and United States 61

ASZ1 Algeria Bahrain Indonesia Jordan Pakistan Qatar Saudi Arabia Egypt United ArabEmirates and Kuwait 40

ASZ3 India 40OTHERS Rest 80

Table 5 Project types and number of responses

CODE PROJECT TYPE NUMBER OF RESPONDENTSISIC1 Information and communication 365ISIC2 Financial and insurance activities 43ISIC3 Construction 42ISIC4 Manufacturing 41ISIC5 Professional scientific and technical activities 32ISIC6 Public administration and defence 12ISIC7 Mining and quarrying 11ISIC8 Human health and social work activities 10ISIC9 Electricity gas steam and air conditioning supply 10ISIC10 Accommodation and food service activities 8ISIC11 Education 8ISIC12 Administrative and support service activities 7ISIC13 Transportation and storage 7ISIC14 Wholesale and retail trade 6ISIC15 Real estate activities 4ISIC16 Development aid 3ISIC17 Arts entertainment and recreation 2

Table 6 Intervals and their interval factor

Interval Meaning Factor (intervalarithmetic mean)

0ndash25 SeldomUnlikely to occur 012525ndash50 OccasionalLikely to occur 037550ndash75 FrequentVery likely to occur 062575ndash100 Always occurs 0875

in such a way that each group is homogeneous and differentfrom the rest For this purposemany data analysis techniquescan be used In this study the Self-organized Maps (SOM)technique has been used a specific type of neural network[60]

SOM networks are an excellent tool for exploring andanalyzing data which are especially adequate due to theirremarkable visualization properties They create a series ofprototype vectors which represent the data set and projectsuch vectors into a low dimensional network (generally

bidimensional) from the input 119889-dimensional space whichpreserves its topology and maintains it As a result thenetwork shows the distance between the different sets sothat it can be used as an adequate visualization surface todisplay different data characteristics as for instance theircluster division Summarizing SOM Clustering Networksallow input data clustering and easily visualize the resultingmultidimensional data clusters For the analysis of datacollected in the questionnaire SOM Toolbox was used withMATLAB [61]Themethodology used is a two-level approachfor partitive clustering where the data set is first projectedusing the SOM and then the SOM is clustered as describedin [62] Partitive clustering algorithms divide a data set intoa number of clusters typically by trying to minimize somecriterion or error function An example of a commonly usedpartitive algorithm is the 119896-means which minimizes errorfunction (2) where 119862 is the number of clusters and 119888119896 is thecenter of cluster 119896

119864 = 119862sum119896=1

sum119909∁119876119896

1003817100381710038171003817119909 minus 11988811989610038171003817100381710038172 (2)

8 Complexity

Table 7 Summary of results

RANK CODE FAILURE CAUSE FREQUENCY INDEX1 C3 Customerrsquos requirements inaccurate incomplete or not defined 059922 C2 Continuous or dramatic changes to initial requirements 056653 C6 Inaccurate time estimations 053294 C16 Project requirements inadequately documented 05092

5 C10 Not or badly defined specifications at the time the Project Team startsto work 04859

6 C5 Inaccurate costs estimations 048267 C8 Lack of Management support 046058 C25 Unrealistic Customerrsquos expectations 04507

9 C4 Disagreements or conflicts of interest among different departmentsinvolved 04478

10 C19 Project Teamrsquos misunderstandings related to CustomerUserrsquos wishesor needs 04417

11 C17 Project staff changes 0430612 C22 Quality checks not or badly performed 0427013 C26 Wrong number of people assigned to the project (too many too few) 0413514 C13 Project Managerrsquos lack of communication skills 0411415 C7 Inadequate management of suppliers and procurement 0382816 C18 Project Teamrsquos lack of competence 0381517 C15 Project Managerrsquos lack of vision 0374618 C14 Project Managerrsquos lack of competence 0374219 C23 Too much complex or new technology 0357420 C20 Project Teamrsquos lack of commitment 0353721 C24 Unexpected events with no effective response possible 0329222 C9 Lack of previous identification of relevant rules and legislation 0319423 C12 Project Managerrsquos lack of commitment 0300924 C11 Political social economic or legal changes 0266225 C21 Public opinion opposition to project 0242826 C1 Competitors 02261

To select the best one among different partitionings eachof these can be evaluated using some kind of validity indexSeveral indices have been proposed [63 64] In our work weused the Davies-Bouldin index [65] which has been provento be suitable for evaluation of 119896-means partitioning Thisindex is a function of the ratio of the sum of within-clusterscatter to between-cluster separation According to Davies-Bouldin validity index the best clustering minimizes (3)which uses 119878119888 for within-cluster distance (119878(119876119896)) and 119889119888119890 forbetween-cluster distance (119889(119876119896 119876119897))

1119862119862sum119896=1

max119897 =119896

119878119888 (119876119896) + 119878119888 (119876119897)119889119888119890 (119876119896 119876119897) (3)

The approach used in this paper (clustering the SOMrather than clustering the data directly) is depicted inFigure 2 First a large set of prototypes (much larger thanthe expected number of clusters) is formed using SOM Theprototypes can be interpreted as ldquoprotoclustersrdquo [62] whichare in the next step combined to form the actual clusters Eachdata vector of the original data set belongs to the same clusteras its nearest prototype

The SOM consists of a regular usually two-dimensionalgrid of map units Each unit 119894 is represented by a prototypevector 119898119894 = [1198981198941 119898119894119889] where 119889 is input vector dimen-sion The units are connected to adjacent ones by a neigh-borhood relation The number of map units which typicallyvaries from a few dozen up to several thousand determinesthe accuracy and generalization capability of the SOM Dur-ing training the SOM forms an elastic net that folds onto theldquocloudrdquo formed by the input data Data points lying near eachother in the input space are mapped onto nearby map unitsThus the SOM can be interpreted as a topology preservingmapping from input space onto 2D grid of map units

The SOM is trained iteratively At each training step asample vector 119909 is randomly chosen from the input data setDistances between 119909 and all the prototype vectors are com-putedThe best-matching unit (BMU) which is denoted hereby 119887 is the map unit with prototype closest to 1199091003817100381710038171003817119909 minus 1198981198871003817100381710038171003817 = min

1198941003817100381710038171003817119909 minus 1198981198941003817100381710038171003817 (4)

Next the prototype vectors are updatedThe BMU and itstopological neighbors are moved closer to the input vector in

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 3

of technological complexity and organizational complexityComplexity can take various forms namely social technolog-ical environmental and organizational Worth mentioninghere is the work of Bosch-Rekveldt et al [19] who proposedthe TOE framework consisting of fifty factors in threefamilies technical organizational and environmental Theauthors have also concluded that organizational complexityworried project managers more than technical or environ-mental complexities Vidal andMarle [4] argued that approx-imately 70 of project complexity factors are organizationalWe follow these assertions by focusing our study mainly onthe organizational factors knowing that this covers a majorarea of complexity in projects

Scholars have focused on the identification of complexityattributes more than any other topic in the field of projectcomplexity Studies in this area have evolved significantly overthe past twenty years Cicmil et al [20] identified complexityas a factor that helps determine planning and control prac-tices and hinders the identification of goals and objectives ora factor that influences time cost and quality of a projectGidado [2] defined project complexity and identified thefactors that influence its effect on project success Also thestudy proposes an approach that measures the complexity ofthe production process in construction

Project complexity has been analyzed in the book editedby Cooke-Davies [21] from three different perspectives peo-ple who manage programs and projects (practitioners) linemanagers in organizations to which programs and projectsmake a substantial contribution (managers) andmembers ofthe academic research community who have an interest inhow complexity shapes and influences the practice of pro-gram and project management (researchers) The book con-stitutes a valuable resource to put together what is currentlyknown and understood about the topic to help practitionersand their managers improve future practice and to guideresearch into answering those questions that will best help toimprove understanding of the topic

Althoughmost authors emphasized the influence of inter-dependencies and interactions of various elements on projectcomplexity [22] few works analyze those dependencies Forexample TOE model [19] does not allow for an understand-ing of how various elements contribute to overall complexityNevertheless other authors regarded project complexity ashaving nonlinear highly dynamic and emerging featuresVidal et al [23] for example proposed the definition ofproject complexity as ldquothe property of a project which makesit difficult to understand foresee and keep under controlits overall behavior even when given reasonably completeinformation about the project systemrdquo Lu et al [24] proposethat project complexity can be defined as ldquoconsisting of manyvaried interrelated parts and has dynamic and emergingfeaturesrdquo Lessard et al built the House of Project Complexity(HoPC) a combined structural and process-based theoreticalframework for understanding contributors to complexity

The connection between failure and complexity in projectmanagement has also been established in the literature Ivoryand Alderman [22] studied the failure in complex systemsin order to shed some critical light on the management ofcomplex projects The authors conclude that it was not the

technical complexity per se that made the management ofthe projects complex Rather the primary determinants ofthe complexity of the project management process stemmedfrom changes in the markets regulatory context and knowl-edge requirements facing the project

There are connections between the project complexityindicators identified in the literature and the project failurecauses For example Kermanshachi et al [3] identified 37indicators Among them several examples usually foundin the literature as project failure causes are included (ieimpact of external agencies on the project execution planimpact of required approvals from external stakeholderslevel of difficulty in obtaining permits level of project designchanges derived by Request for Information (RFI) etc) Infact there are several examples of authors that are referencedfrequently in the field of both project complexity and projectsuccess factors Shenhar et al [25] Dvir and Lechler [26]Cooke-Davies [27] and Pinto and Prescott [28] are someexamples of those

22 Project SuccessFailure Factors One of the most relevantfields of study in project management is success factors andfailure causes in projectsTheywere first identified at the PMIAnnual Seminar and Symposium in 1986 [29] and becameone of themost discussed themeswithin specialized literature[30] Each stakeholder working on a project has hisher ownopinion on what is determinant for failure and it is muchmore complex than adhering to the traditional criteria oftime cost and quality If a project can be considered failedwhen it has not delivered what was expected after its com-pletion causes leading to this unsatisfactory result are alsosubjected to the different points of view of the stakeholdersinvolved

The different lists of success factors and failure causesin the literature show no consensus The most usual wouldbe a combination of several factors with different levels ofinfluence in different stages of the projectrsquos life cycle resultingin its success or failure The concepts of success factors andfailure causes are closely related but a failure cause is notnecessarily the negation of or the opposite to a successfactorThere is not always such correspondence among themThis study is considering failure causes rather than successfulfactors because we are taking the assumption that the morethe failure causes concur in a project the higher the complex-ity is

The literature identifying successfailure factors is veryextensive Some examples especially regarding failure causesare [31ndash35] all of them related to the construction sectorAnother sector with many references is the IT (informationtechnology) projects Some examples are [36ndash39] There areseveral frameworks classifying the factors Belassi and Tukel[40] suggested a scheme that classifies the critical factorsin four different dimensions (Table 1) and describes theimpacts of these factors on project performance Shenhar etal [25] have identified also four distinct dimensions projectefficiency impact on the customer direct and business suc-cess and preparing for the future They stated that the exactcontent of each dimension and its relative importancemay change with time and are contingent on the specific

4 Complexity

Table 1 Belassi and Tukel taxonomy

Groups of factorsRelated to projectRelated to the project managers and the team membersFactors related to the organizationFactors related to the external environment

stakeholder Lim and Mohamed [41] consider two view-points of project success macro and micro Regarding themicro they have identified technical commercial financerisk environmental and human related factors

Richardson [42] and King [43] point out that none of thekey success factors described in the literature is responsibleon its own for ensuring a projectrsquos success They are allinterdependent and require a holistic approach Groups ofsuccess factors and their interactions are of prime importancein determining a projectrsquos success or failure [44] Multivariatestatisticsmethodsmay be very useful with this purpose Someexamples applied in the project management complexity are[4 45 46] as well as in the field of project success [47ndash49]Clustering methods have been also used in the context ofexploring complex relationships within the field of projectmanagement for example the works [50ndash52] Our study usesself-organizing maps (SOM) and clustering techniques tofind patterns in a data set of answers coming from a survey

23 Self-Organizing Maps and Applications in Project Man-agement SOM is an unsupervised neural network proposedby Kohonen [53] for visual cluster analysis The neurons ofthe map are located on a regular grid embedded in a low(usually 2 or 3) dimensional space and associated with thecluster prototypes by the connected weights In the course ofthe learning process the neurons compete with each otherthrough the best-matching principle in such a way that theinput is projected to the nearest neuron given a defineddistance metric The winner neuron and its neighbors on themap are then adjusted towards the input in proportion withthe neighborhood distance consequently the neighboringneurons likely represent similar patterns of the input dataspace Due to the data clustering and spatialization throughthe topology preserving projection SOM is widely usedin the context of visual clustering applications Despite theunsupervised nature the applicability of SOM is extendedto classification tasks by means of a variety of ways suchas neuron labeling method semisupervised learning orsupervised learning vector quantization (LVQ) [54]

SOM is recognized as a useful technique to analyze high-dimensional data sets and understand their hidden relationsIt can be used to manage complexity in large data sets [39]Nevertheless there are few cases described in the field ofproject management Balsera et al [55] have exposed theapplication of SOM to analyze information related to effortestimation and software projects features MacDonell [56]has reported other multidimensional data study visualizationbased on SOM identifying groups of data for similar projectsand finding nonlinear relationships within the exploredvariables Naiem et al [57] have used SOM for visualizing the

Table 2 Failure causes considered in the study

Cause CodeIrruption of competitors C1Continuous or dramatic changes to the initial requirements C2Customerrsquos requirements inaccurate incomplete or notdefined C3

Disagreement or conflict of interest among departments C4Inaccurate cost estimations C5Inaccurate time estimations C6Deficient management of suppliers and procurement C7Lack of Management support C8Lack of previous identification of legislation C9Badly defined specifications C10Political social economic or legal changes C11Project manager Lack of commitment C12Project manager Lack of communication skills C13Project manager Lack of competence C14Project manager Lack of vision C15Project requirements deficiently documented C16Project staff changes C17Project team lack of competence C18Project team misunderstanding related to customeruserneeds C19

Project team lack of commitment C20Public opinion opposition to Project C21Quality checks badly performed or not performed at all C22Extremely new or complex technology C23Unexpected events with no effective response possible C24Unrealistic customer expectations C25Wrong number of people assigned to the project C26

set of candidate portfolios After reviewing the literature wecan conclude that there is not any former application of SOMfor analyzing patterns of project failure causes

3 Research Methodology

As discussed in the Introduction the ground for this workis the questionnaire that was designed to gather informationon the perception project managers have of what the successfactors and failure causes are After the information wasgathered a multivariate analysis was performed on the datawith cluster data mining techniques

The sections of the questionnaire considered for thisstudy were as follows

(i) General information on the respondent and typologyof projects heshe was involved in country type andsize of project

(ii) Frequency of different causes of project failure with26 multiple-choice questions (from 0ndash25 rare orimprobable occurrence to 75ndash100 always occurs)

The causes of failure are extracted from the existingbibliography and the previous work on the matter Table 2

Complexity 5

Factors related to organization

middotDisagreement or conflict of interest among departments (C4)

middotDeficient management of suppliers and procurement (C7)

middotLack of Management support (C8)

middotProject staff changes (C17)

middotWrong number of people assigned to the project (C26)

Factors related to project

middotContinuous or dramatic changes to the initial requirements (C2)middotCustomerrsquos requirements inaccurate incomplete or not defined (C3)middotInaccurate cost estimations (C5)middotInaccurate time estimations (C6)middotLack of previous identification of legislation (C9)-Badly defined specifications (C10)middotProject requirements deficiently documented (C16)middotQuality checks badly performed or not performed at all (C22)middotExtremely new or complex technology (C23)

Factors related to the project manager and team members

middotProject manager Lack of commitment (C12)

middotProject manager Lack of communication skills (C13)

middotProject manager Lack of competence (C14)

middotProject manager Lack of vision (C15)

middotProject team lack of competence (C18)

middotProject team misunderstanding related to customeruser needs (C19)

middotProject team lack of commitment (C20)

Factors related to external environment

middotIrruption of competitors (C1)

middotPolitical social economic or legal changes (C11)

middotPublic opinion opposition to Project (C21)

middotUnexpected events with no effective response possible (C24)

middotUnrealistic customer expectations (C25)

FAILURE

Figure 1 Failure causes grouped by category The factors are derived from existing bibliography and previous work on the matter They havebeen classified according to Belassi and Tukel taxonomy

gathers the identified factors The factors are coded as 119862119909where 119909 is the number given to the failure cause

The factors were classified according to Belassi andTukel taxonomy [40] included in Table 1 Figure 1 depictsthe classification considered The factors are presented un-shorted because they were shuffled in order to avoid biasesin the survey The first group includes the factors related tothe organization the project belongs to (ie factors relatedto top management support) The second group includes thefactors related to the project itself and the way the projectis managed The third one comprises the factors related tothe project manager and the team members Many studiesdemonstrated the importance of selecting project managerswho possess the necessary technical and administrativeskills for successful project termination as well as theircommitmentThe competences of the teammembers are alsofound to be a critical factor Finally the factors related toexternal environment consist of factors which are external tothe organization but still have an impact on project successor failure The classification can be considered collectivelyexhaustive Belassi and Tukel state that the four groupsoffer a comprehensive set in that any factor listed in the

literature or even specific points of consideration shouldbelong to at least one group [40] However it is not easyto differentiate in which category to include each of thedifferent factors The groups are interrelated The authorsgive some indications to distinguish where to classify themmainly regarding organization and external environmentFor example if a customer is from outside the organizationheshe should be considered as an external factor (ie C25-unrealistic customer expectations) For functional projectshowever customers are usually part of the organization suchas topmanagement In such cases factors related to the clientcan be grouped under the factors related to the organization

We can find correspondences among most of the failurecauses included in our study (Table 2) and the complexityindicators appointed by Kermanshachi et al [3] For exampleC2 (which is one of the most frequent in accordance withthe rank presented in Table 7) is related to several complexityindexes as ldquolevel of project design changes derived by RFIrdquoand ldquomagnitude of change ordersrdquo C10 for example isrelated to ldquopercentage of design completed at the start ofconstructionrdquo This fact reinforces the connection betweenfailure causes and complexity factors

6 Complexity

The recipients were randomly chosen among the mem-bers of 36 project management groups from LinkedIn net-work The questionnaire was open for 3 months and 11 daysin 2011 in order to obtain a significant number of answersDuring that period of time customized emails were sentwith an invitation to answer the questionnaire to a total of3668 people A total of 619 answers were received (1688)611 of which were considered for further analysis (the restwere discarded due to consistency issues) Neither of thequestionnaires was partially filled or incomplete since allfields were marked as mandatory Previously in 2010 a pilotsurvey was conducted with the help of project managementexperts and practitioners This primary questionnaire wassent up to 45 people with a response rate of 6667 Thosefactors which reached a higher score were the ones finallyselected to configure the list included in the current studySuggestions and comments made by respondents to improvethe proposed list were also taken into account

Answers from 63 countries were received as Table 3shows

The answers have been grouped first into 13 differentgeographical zones (more details can be found in [16])according to geographical cultural historical and economiccriteria From those a total of 6 groups comprise more than85 of the respondents AMZ3 EUZ3 EUZ1 AMZ1 ASZ1and ASZ3 The other 15 of the responses were grouped in anew category called ldquoOthersrdquo Table 4 summarizes the groupsby geographical zone

Following a similar approach 53 types of projects werepresent on the answers received To simplify they have beenclassified into a total of 17 groups according to the highestlevel of the ISIC Rev 4 Classification [58] Development aidprojects were listed apart as an independent category Table 5presents the number of respondents from each project typeand the codification

Information about the size of the projects the respondentsare usually involved in was also enquired The size of theprojects is something quite subjective anddifficult to compareamong different sectors For example considering the projectbudget as an indicator of project size a project consideredsmall in construction could by contrast be consideredlarge in information technology projects In order to avoidthese biases the guideline provided by the University ofWisconsin-Madison [59] was used The results show that1195 of respondents are involved in small projects 4926in medium size projects and 3879 in large projects

Though the scope of this paper does not cover thestatistical analysis of the obtained answers here is a summaryof the main results Overall both the most and least frequentfailure causes are presented here In order to rank each one ofthe 26 failure causes a frequency index (FI) was calculated asfollows

FI = [sum611119894=1 sum4119895=1 119860 119894 lowast 119861119895]611 (1)

where119860 119894 is the number of responses choosing interval 119894119861119895 is interval factor 119895

Table 3 Countries and number of responses (ordered by numberof responses) When several countries are indicated in the samerow the number of respondents stands for the number of answersreceived from each country (ie 50 answers from Spain and 50answers from the United States)

COUNTRIES NUMBER OFRESPONDENTS

Argentina 91Spain United States 50Greece 45Chile 41India 40Brazil 30Luxembourg 27Mexico 21Uruguay 20United Arab Emirates 19Italy 18Russia 16Sweden 14South Africa 12Canada 11Germany 8Saudi Arabia The Netherlands UnitedKingdom 7

Australia Denmark Finland Kuwait 5Belgium Paraguay Slovenia 4Egypt France Israel Switzerland 3Philippines Tanzania 2Algeria Austria Bahrain Belarus BotswanaBritish Virgin Islands Cameroon ChinaColombia Cyprus Czech Republic GhanaHong Kong Hungary Indonesia IrelandJapan Jordan Nigeria Pakistan Peru PolandQatar Rwanda Singapore Surinam ThailandUganda Venezuela

1

Each interval factor is defined as presented in Table 6 Theranking is presented in Table 7

With the obtained results a cluster analysis was con-ducted to obtain patterns in the answers data set groupinga set of data with similar values The aim is to classify a set ofsimple elements into a number of groups in such a way thatthe elements in the same group are similar or related to oneanother and at the same time different or unrelated to theelements in other groups

The cluster analysis (also called Unsupervised Classi-fication Exploratory Data Analysis Clustering NumericalTaxonomy or PatternRecognition) is amultivariate statisticaltechnique whose aim is to divide a set of objects into groupsor clusters in such a way that objects in the same cluster arevery similar to each other (internal cluster cohesion) andthe objects in other groups are different (external clusterisolation) Summarizing it deals with creating data clusters

Complexity 7

Table 4 Answers grouped by geographical zone

GEOGRAPHICAL ZONESANALYZED COUNTRIES NUMBER OF

RESPONDENTSAMZ3 Argentina Brazil Chile Colombia Paraguay Peru Surinam Uruguay and Venezuela 190EUZ3 Cyprus Spain Greece and Italy 114

EUZ1 Germany Austria Belgium Denmark Finland France Ireland Virgin Islands (UK)Luxembourg United Kingdom The Netherlands Sweden and Switzerland 86

AMZ1 Canada and United States 61

ASZ1 Algeria Bahrain Indonesia Jordan Pakistan Qatar Saudi Arabia Egypt United ArabEmirates and Kuwait 40

ASZ3 India 40OTHERS Rest 80

Table 5 Project types and number of responses

CODE PROJECT TYPE NUMBER OF RESPONDENTSISIC1 Information and communication 365ISIC2 Financial and insurance activities 43ISIC3 Construction 42ISIC4 Manufacturing 41ISIC5 Professional scientific and technical activities 32ISIC6 Public administration and defence 12ISIC7 Mining and quarrying 11ISIC8 Human health and social work activities 10ISIC9 Electricity gas steam and air conditioning supply 10ISIC10 Accommodation and food service activities 8ISIC11 Education 8ISIC12 Administrative and support service activities 7ISIC13 Transportation and storage 7ISIC14 Wholesale and retail trade 6ISIC15 Real estate activities 4ISIC16 Development aid 3ISIC17 Arts entertainment and recreation 2

Table 6 Intervals and their interval factor

Interval Meaning Factor (intervalarithmetic mean)

0ndash25 SeldomUnlikely to occur 012525ndash50 OccasionalLikely to occur 037550ndash75 FrequentVery likely to occur 062575ndash100 Always occurs 0875

in such a way that each group is homogeneous and differentfrom the rest For this purposemany data analysis techniquescan be used In this study the Self-organized Maps (SOM)technique has been used a specific type of neural network[60]

SOM networks are an excellent tool for exploring andanalyzing data which are especially adequate due to theirremarkable visualization properties They create a series ofprototype vectors which represent the data set and projectsuch vectors into a low dimensional network (generally

bidimensional) from the input 119889-dimensional space whichpreserves its topology and maintains it As a result thenetwork shows the distance between the different sets sothat it can be used as an adequate visualization surface todisplay different data characteristics as for instance theircluster division Summarizing SOM Clustering Networksallow input data clustering and easily visualize the resultingmultidimensional data clusters For the analysis of datacollected in the questionnaire SOM Toolbox was used withMATLAB [61]Themethodology used is a two-level approachfor partitive clustering where the data set is first projectedusing the SOM and then the SOM is clustered as describedin [62] Partitive clustering algorithms divide a data set intoa number of clusters typically by trying to minimize somecriterion or error function An example of a commonly usedpartitive algorithm is the 119896-means which minimizes errorfunction (2) where 119862 is the number of clusters and 119888119896 is thecenter of cluster 119896

119864 = 119862sum119896=1

sum119909∁119876119896

1003817100381710038171003817119909 minus 11988811989610038171003817100381710038172 (2)

8 Complexity

Table 7 Summary of results

RANK CODE FAILURE CAUSE FREQUENCY INDEX1 C3 Customerrsquos requirements inaccurate incomplete or not defined 059922 C2 Continuous or dramatic changes to initial requirements 056653 C6 Inaccurate time estimations 053294 C16 Project requirements inadequately documented 05092

5 C10 Not or badly defined specifications at the time the Project Team startsto work 04859

6 C5 Inaccurate costs estimations 048267 C8 Lack of Management support 046058 C25 Unrealistic Customerrsquos expectations 04507

9 C4 Disagreements or conflicts of interest among different departmentsinvolved 04478

10 C19 Project Teamrsquos misunderstandings related to CustomerUserrsquos wishesor needs 04417

11 C17 Project staff changes 0430612 C22 Quality checks not or badly performed 0427013 C26 Wrong number of people assigned to the project (too many too few) 0413514 C13 Project Managerrsquos lack of communication skills 0411415 C7 Inadequate management of suppliers and procurement 0382816 C18 Project Teamrsquos lack of competence 0381517 C15 Project Managerrsquos lack of vision 0374618 C14 Project Managerrsquos lack of competence 0374219 C23 Too much complex or new technology 0357420 C20 Project Teamrsquos lack of commitment 0353721 C24 Unexpected events with no effective response possible 0329222 C9 Lack of previous identification of relevant rules and legislation 0319423 C12 Project Managerrsquos lack of commitment 0300924 C11 Political social economic or legal changes 0266225 C21 Public opinion opposition to project 0242826 C1 Competitors 02261

To select the best one among different partitionings eachof these can be evaluated using some kind of validity indexSeveral indices have been proposed [63 64] In our work weused the Davies-Bouldin index [65] which has been provento be suitable for evaluation of 119896-means partitioning Thisindex is a function of the ratio of the sum of within-clusterscatter to between-cluster separation According to Davies-Bouldin validity index the best clustering minimizes (3)which uses 119878119888 for within-cluster distance (119878(119876119896)) and 119889119888119890 forbetween-cluster distance (119889(119876119896 119876119897))

1119862119862sum119896=1

max119897 =119896

119878119888 (119876119896) + 119878119888 (119876119897)119889119888119890 (119876119896 119876119897) (3)

The approach used in this paper (clustering the SOMrather than clustering the data directly) is depicted inFigure 2 First a large set of prototypes (much larger thanthe expected number of clusters) is formed using SOM Theprototypes can be interpreted as ldquoprotoclustersrdquo [62] whichare in the next step combined to form the actual clusters Eachdata vector of the original data set belongs to the same clusteras its nearest prototype

The SOM consists of a regular usually two-dimensionalgrid of map units Each unit 119894 is represented by a prototypevector 119898119894 = [1198981198941 119898119894119889] where 119889 is input vector dimen-sion The units are connected to adjacent ones by a neigh-borhood relation The number of map units which typicallyvaries from a few dozen up to several thousand determinesthe accuracy and generalization capability of the SOM Dur-ing training the SOM forms an elastic net that folds onto theldquocloudrdquo formed by the input data Data points lying near eachother in the input space are mapped onto nearby map unitsThus the SOM can be interpreted as a topology preservingmapping from input space onto 2D grid of map units

The SOM is trained iteratively At each training step asample vector 119909 is randomly chosen from the input data setDistances between 119909 and all the prototype vectors are com-putedThe best-matching unit (BMU) which is denoted hereby 119887 is the map unit with prototype closest to 1199091003817100381710038171003817119909 minus 1198981198871003817100381710038171003817 = min

1198941003817100381710038171003817119909 minus 1198981198941003817100381710038171003817 (4)

Next the prototype vectors are updatedThe BMU and itstopological neighbors are moved closer to the input vector in

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

4 Complexity

Table 1 Belassi and Tukel taxonomy

Groups of factorsRelated to projectRelated to the project managers and the team membersFactors related to the organizationFactors related to the external environment

stakeholder Lim and Mohamed [41] consider two view-points of project success macro and micro Regarding themicro they have identified technical commercial financerisk environmental and human related factors

Richardson [42] and King [43] point out that none of thekey success factors described in the literature is responsibleon its own for ensuring a projectrsquos success They are allinterdependent and require a holistic approach Groups ofsuccess factors and their interactions are of prime importancein determining a projectrsquos success or failure [44] Multivariatestatisticsmethodsmay be very useful with this purpose Someexamples applied in the project management complexity are[4 45 46] as well as in the field of project success [47ndash49]Clustering methods have been also used in the context ofexploring complex relationships within the field of projectmanagement for example the works [50ndash52] Our study usesself-organizing maps (SOM) and clustering techniques tofind patterns in a data set of answers coming from a survey

23 Self-Organizing Maps and Applications in Project Man-agement SOM is an unsupervised neural network proposedby Kohonen [53] for visual cluster analysis The neurons ofthe map are located on a regular grid embedded in a low(usually 2 or 3) dimensional space and associated with thecluster prototypes by the connected weights In the course ofthe learning process the neurons compete with each otherthrough the best-matching principle in such a way that theinput is projected to the nearest neuron given a defineddistance metric The winner neuron and its neighbors on themap are then adjusted towards the input in proportion withthe neighborhood distance consequently the neighboringneurons likely represent similar patterns of the input dataspace Due to the data clustering and spatialization throughthe topology preserving projection SOM is widely usedin the context of visual clustering applications Despite theunsupervised nature the applicability of SOM is extendedto classification tasks by means of a variety of ways suchas neuron labeling method semisupervised learning orsupervised learning vector quantization (LVQ) [54]

SOM is recognized as a useful technique to analyze high-dimensional data sets and understand their hidden relationsIt can be used to manage complexity in large data sets [39]Nevertheless there are few cases described in the field ofproject management Balsera et al [55] have exposed theapplication of SOM to analyze information related to effortestimation and software projects features MacDonell [56]has reported other multidimensional data study visualizationbased on SOM identifying groups of data for similar projectsand finding nonlinear relationships within the exploredvariables Naiem et al [57] have used SOM for visualizing the

Table 2 Failure causes considered in the study

Cause CodeIrruption of competitors C1Continuous or dramatic changes to the initial requirements C2Customerrsquos requirements inaccurate incomplete or notdefined C3

Disagreement or conflict of interest among departments C4Inaccurate cost estimations C5Inaccurate time estimations C6Deficient management of suppliers and procurement C7Lack of Management support C8Lack of previous identification of legislation C9Badly defined specifications C10Political social economic or legal changes C11Project manager Lack of commitment C12Project manager Lack of communication skills C13Project manager Lack of competence C14Project manager Lack of vision C15Project requirements deficiently documented C16Project staff changes C17Project team lack of competence C18Project team misunderstanding related to customeruserneeds C19

Project team lack of commitment C20Public opinion opposition to Project C21Quality checks badly performed or not performed at all C22Extremely new or complex technology C23Unexpected events with no effective response possible C24Unrealistic customer expectations C25Wrong number of people assigned to the project C26

set of candidate portfolios After reviewing the literature wecan conclude that there is not any former application of SOMfor analyzing patterns of project failure causes

3 Research Methodology

As discussed in the Introduction the ground for this workis the questionnaire that was designed to gather informationon the perception project managers have of what the successfactors and failure causes are After the information wasgathered a multivariate analysis was performed on the datawith cluster data mining techniques

The sections of the questionnaire considered for thisstudy were as follows

(i) General information on the respondent and typologyof projects heshe was involved in country type andsize of project

(ii) Frequency of different causes of project failure with26 multiple-choice questions (from 0ndash25 rare orimprobable occurrence to 75ndash100 always occurs)

The causes of failure are extracted from the existingbibliography and the previous work on the matter Table 2

Complexity 5

Factors related to organization

middotDisagreement or conflict of interest among departments (C4)

middotDeficient management of suppliers and procurement (C7)

middotLack of Management support (C8)

middotProject staff changes (C17)

middotWrong number of people assigned to the project (C26)

Factors related to project

middotContinuous or dramatic changes to the initial requirements (C2)middotCustomerrsquos requirements inaccurate incomplete or not defined (C3)middotInaccurate cost estimations (C5)middotInaccurate time estimations (C6)middotLack of previous identification of legislation (C9)-Badly defined specifications (C10)middotProject requirements deficiently documented (C16)middotQuality checks badly performed or not performed at all (C22)middotExtremely new or complex technology (C23)

Factors related to the project manager and team members

middotProject manager Lack of commitment (C12)

middotProject manager Lack of communication skills (C13)

middotProject manager Lack of competence (C14)

middotProject manager Lack of vision (C15)

middotProject team lack of competence (C18)

middotProject team misunderstanding related to customeruser needs (C19)

middotProject team lack of commitment (C20)

Factors related to external environment

middotIrruption of competitors (C1)

middotPolitical social economic or legal changes (C11)

middotPublic opinion opposition to Project (C21)

middotUnexpected events with no effective response possible (C24)

middotUnrealistic customer expectations (C25)

FAILURE

Figure 1 Failure causes grouped by category The factors are derived from existing bibliography and previous work on the matter They havebeen classified according to Belassi and Tukel taxonomy

gathers the identified factors The factors are coded as 119862119909where 119909 is the number given to the failure cause

The factors were classified according to Belassi andTukel taxonomy [40] included in Table 1 Figure 1 depictsthe classification considered The factors are presented un-shorted because they were shuffled in order to avoid biasesin the survey The first group includes the factors related tothe organization the project belongs to (ie factors relatedto top management support) The second group includes thefactors related to the project itself and the way the projectis managed The third one comprises the factors related tothe project manager and the team members Many studiesdemonstrated the importance of selecting project managerswho possess the necessary technical and administrativeskills for successful project termination as well as theircommitmentThe competences of the teammembers are alsofound to be a critical factor Finally the factors related toexternal environment consist of factors which are external tothe organization but still have an impact on project successor failure The classification can be considered collectivelyexhaustive Belassi and Tukel state that the four groupsoffer a comprehensive set in that any factor listed in the

literature or even specific points of consideration shouldbelong to at least one group [40] However it is not easyto differentiate in which category to include each of thedifferent factors The groups are interrelated The authorsgive some indications to distinguish where to classify themmainly regarding organization and external environmentFor example if a customer is from outside the organizationheshe should be considered as an external factor (ie C25-unrealistic customer expectations) For functional projectshowever customers are usually part of the organization suchas topmanagement In such cases factors related to the clientcan be grouped under the factors related to the organization

We can find correspondences among most of the failurecauses included in our study (Table 2) and the complexityindicators appointed by Kermanshachi et al [3] For exampleC2 (which is one of the most frequent in accordance withthe rank presented in Table 7) is related to several complexityindexes as ldquolevel of project design changes derived by RFIrdquoand ldquomagnitude of change ordersrdquo C10 for example isrelated to ldquopercentage of design completed at the start ofconstructionrdquo This fact reinforces the connection betweenfailure causes and complexity factors

6 Complexity

The recipients were randomly chosen among the mem-bers of 36 project management groups from LinkedIn net-work The questionnaire was open for 3 months and 11 daysin 2011 in order to obtain a significant number of answersDuring that period of time customized emails were sentwith an invitation to answer the questionnaire to a total of3668 people A total of 619 answers were received (1688)611 of which were considered for further analysis (the restwere discarded due to consistency issues) Neither of thequestionnaires was partially filled or incomplete since allfields were marked as mandatory Previously in 2010 a pilotsurvey was conducted with the help of project managementexperts and practitioners This primary questionnaire wassent up to 45 people with a response rate of 6667 Thosefactors which reached a higher score were the ones finallyselected to configure the list included in the current studySuggestions and comments made by respondents to improvethe proposed list were also taken into account

Answers from 63 countries were received as Table 3shows

The answers have been grouped first into 13 differentgeographical zones (more details can be found in [16])according to geographical cultural historical and economiccriteria From those a total of 6 groups comprise more than85 of the respondents AMZ3 EUZ3 EUZ1 AMZ1 ASZ1and ASZ3 The other 15 of the responses were grouped in anew category called ldquoOthersrdquo Table 4 summarizes the groupsby geographical zone

Following a similar approach 53 types of projects werepresent on the answers received To simplify they have beenclassified into a total of 17 groups according to the highestlevel of the ISIC Rev 4 Classification [58] Development aidprojects were listed apart as an independent category Table 5presents the number of respondents from each project typeand the codification

Information about the size of the projects the respondentsare usually involved in was also enquired The size of theprojects is something quite subjective anddifficult to compareamong different sectors For example considering the projectbudget as an indicator of project size a project consideredsmall in construction could by contrast be consideredlarge in information technology projects In order to avoidthese biases the guideline provided by the University ofWisconsin-Madison [59] was used The results show that1195 of respondents are involved in small projects 4926in medium size projects and 3879 in large projects

Though the scope of this paper does not cover thestatistical analysis of the obtained answers here is a summaryof the main results Overall both the most and least frequentfailure causes are presented here In order to rank each one ofthe 26 failure causes a frequency index (FI) was calculated asfollows

FI = [sum611119894=1 sum4119895=1 119860 119894 lowast 119861119895]611 (1)

where119860 119894 is the number of responses choosing interval 119894119861119895 is interval factor 119895

Table 3 Countries and number of responses (ordered by numberof responses) When several countries are indicated in the samerow the number of respondents stands for the number of answersreceived from each country (ie 50 answers from Spain and 50answers from the United States)

COUNTRIES NUMBER OFRESPONDENTS

Argentina 91Spain United States 50Greece 45Chile 41India 40Brazil 30Luxembourg 27Mexico 21Uruguay 20United Arab Emirates 19Italy 18Russia 16Sweden 14South Africa 12Canada 11Germany 8Saudi Arabia The Netherlands UnitedKingdom 7

Australia Denmark Finland Kuwait 5Belgium Paraguay Slovenia 4Egypt France Israel Switzerland 3Philippines Tanzania 2Algeria Austria Bahrain Belarus BotswanaBritish Virgin Islands Cameroon ChinaColombia Cyprus Czech Republic GhanaHong Kong Hungary Indonesia IrelandJapan Jordan Nigeria Pakistan Peru PolandQatar Rwanda Singapore Surinam ThailandUganda Venezuela

1

Each interval factor is defined as presented in Table 6 Theranking is presented in Table 7

With the obtained results a cluster analysis was con-ducted to obtain patterns in the answers data set groupinga set of data with similar values The aim is to classify a set ofsimple elements into a number of groups in such a way thatthe elements in the same group are similar or related to oneanother and at the same time different or unrelated to theelements in other groups

The cluster analysis (also called Unsupervised Classi-fication Exploratory Data Analysis Clustering NumericalTaxonomy or PatternRecognition) is amultivariate statisticaltechnique whose aim is to divide a set of objects into groupsor clusters in such a way that objects in the same cluster arevery similar to each other (internal cluster cohesion) andthe objects in other groups are different (external clusterisolation) Summarizing it deals with creating data clusters

Complexity 7

Table 4 Answers grouped by geographical zone

GEOGRAPHICAL ZONESANALYZED COUNTRIES NUMBER OF

RESPONDENTSAMZ3 Argentina Brazil Chile Colombia Paraguay Peru Surinam Uruguay and Venezuela 190EUZ3 Cyprus Spain Greece and Italy 114

EUZ1 Germany Austria Belgium Denmark Finland France Ireland Virgin Islands (UK)Luxembourg United Kingdom The Netherlands Sweden and Switzerland 86

AMZ1 Canada and United States 61

ASZ1 Algeria Bahrain Indonesia Jordan Pakistan Qatar Saudi Arabia Egypt United ArabEmirates and Kuwait 40

ASZ3 India 40OTHERS Rest 80

Table 5 Project types and number of responses

CODE PROJECT TYPE NUMBER OF RESPONDENTSISIC1 Information and communication 365ISIC2 Financial and insurance activities 43ISIC3 Construction 42ISIC4 Manufacturing 41ISIC5 Professional scientific and technical activities 32ISIC6 Public administration and defence 12ISIC7 Mining and quarrying 11ISIC8 Human health and social work activities 10ISIC9 Electricity gas steam and air conditioning supply 10ISIC10 Accommodation and food service activities 8ISIC11 Education 8ISIC12 Administrative and support service activities 7ISIC13 Transportation and storage 7ISIC14 Wholesale and retail trade 6ISIC15 Real estate activities 4ISIC16 Development aid 3ISIC17 Arts entertainment and recreation 2

Table 6 Intervals and their interval factor

Interval Meaning Factor (intervalarithmetic mean)

0ndash25 SeldomUnlikely to occur 012525ndash50 OccasionalLikely to occur 037550ndash75 FrequentVery likely to occur 062575ndash100 Always occurs 0875

in such a way that each group is homogeneous and differentfrom the rest For this purposemany data analysis techniquescan be used In this study the Self-organized Maps (SOM)technique has been used a specific type of neural network[60]

SOM networks are an excellent tool for exploring andanalyzing data which are especially adequate due to theirremarkable visualization properties They create a series ofprototype vectors which represent the data set and projectsuch vectors into a low dimensional network (generally

bidimensional) from the input 119889-dimensional space whichpreserves its topology and maintains it As a result thenetwork shows the distance between the different sets sothat it can be used as an adequate visualization surface todisplay different data characteristics as for instance theircluster division Summarizing SOM Clustering Networksallow input data clustering and easily visualize the resultingmultidimensional data clusters For the analysis of datacollected in the questionnaire SOM Toolbox was used withMATLAB [61]Themethodology used is a two-level approachfor partitive clustering where the data set is first projectedusing the SOM and then the SOM is clustered as describedin [62] Partitive clustering algorithms divide a data set intoa number of clusters typically by trying to minimize somecriterion or error function An example of a commonly usedpartitive algorithm is the 119896-means which minimizes errorfunction (2) where 119862 is the number of clusters and 119888119896 is thecenter of cluster 119896

119864 = 119862sum119896=1

sum119909∁119876119896

1003817100381710038171003817119909 minus 11988811989610038171003817100381710038172 (2)

8 Complexity

Table 7 Summary of results

RANK CODE FAILURE CAUSE FREQUENCY INDEX1 C3 Customerrsquos requirements inaccurate incomplete or not defined 059922 C2 Continuous or dramatic changes to initial requirements 056653 C6 Inaccurate time estimations 053294 C16 Project requirements inadequately documented 05092

5 C10 Not or badly defined specifications at the time the Project Team startsto work 04859

6 C5 Inaccurate costs estimations 048267 C8 Lack of Management support 046058 C25 Unrealistic Customerrsquos expectations 04507

9 C4 Disagreements or conflicts of interest among different departmentsinvolved 04478

10 C19 Project Teamrsquos misunderstandings related to CustomerUserrsquos wishesor needs 04417

11 C17 Project staff changes 0430612 C22 Quality checks not or badly performed 0427013 C26 Wrong number of people assigned to the project (too many too few) 0413514 C13 Project Managerrsquos lack of communication skills 0411415 C7 Inadequate management of suppliers and procurement 0382816 C18 Project Teamrsquos lack of competence 0381517 C15 Project Managerrsquos lack of vision 0374618 C14 Project Managerrsquos lack of competence 0374219 C23 Too much complex or new technology 0357420 C20 Project Teamrsquos lack of commitment 0353721 C24 Unexpected events with no effective response possible 0329222 C9 Lack of previous identification of relevant rules and legislation 0319423 C12 Project Managerrsquos lack of commitment 0300924 C11 Political social economic or legal changes 0266225 C21 Public opinion opposition to project 0242826 C1 Competitors 02261

To select the best one among different partitionings eachof these can be evaluated using some kind of validity indexSeveral indices have been proposed [63 64] In our work weused the Davies-Bouldin index [65] which has been provento be suitable for evaluation of 119896-means partitioning Thisindex is a function of the ratio of the sum of within-clusterscatter to between-cluster separation According to Davies-Bouldin validity index the best clustering minimizes (3)which uses 119878119888 for within-cluster distance (119878(119876119896)) and 119889119888119890 forbetween-cluster distance (119889(119876119896 119876119897))

1119862119862sum119896=1

max119897 =119896

119878119888 (119876119896) + 119878119888 (119876119897)119889119888119890 (119876119896 119876119897) (3)

The approach used in this paper (clustering the SOMrather than clustering the data directly) is depicted inFigure 2 First a large set of prototypes (much larger thanthe expected number of clusters) is formed using SOM Theprototypes can be interpreted as ldquoprotoclustersrdquo [62] whichare in the next step combined to form the actual clusters Eachdata vector of the original data set belongs to the same clusteras its nearest prototype

The SOM consists of a regular usually two-dimensionalgrid of map units Each unit 119894 is represented by a prototypevector 119898119894 = [1198981198941 119898119894119889] where 119889 is input vector dimen-sion The units are connected to adjacent ones by a neigh-borhood relation The number of map units which typicallyvaries from a few dozen up to several thousand determinesthe accuracy and generalization capability of the SOM Dur-ing training the SOM forms an elastic net that folds onto theldquocloudrdquo formed by the input data Data points lying near eachother in the input space are mapped onto nearby map unitsThus the SOM can be interpreted as a topology preservingmapping from input space onto 2D grid of map units

The SOM is trained iteratively At each training step asample vector 119909 is randomly chosen from the input data setDistances between 119909 and all the prototype vectors are com-putedThe best-matching unit (BMU) which is denoted hereby 119887 is the map unit with prototype closest to 1199091003817100381710038171003817119909 minus 1198981198871003817100381710038171003817 = min

1198941003817100381710038171003817119909 minus 1198981198941003817100381710038171003817 (4)

Next the prototype vectors are updatedThe BMU and itstopological neighbors are moved closer to the input vector in

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 5

Factors related to organization

middotDisagreement or conflict of interest among departments (C4)

middotDeficient management of suppliers and procurement (C7)

middotLack of Management support (C8)

middotProject staff changes (C17)

middotWrong number of people assigned to the project (C26)

Factors related to project

middotContinuous or dramatic changes to the initial requirements (C2)middotCustomerrsquos requirements inaccurate incomplete or not defined (C3)middotInaccurate cost estimations (C5)middotInaccurate time estimations (C6)middotLack of previous identification of legislation (C9)-Badly defined specifications (C10)middotProject requirements deficiently documented (C16)middotQuality checks badly performed or not performed at all (C22)middotExtremely new or complex technology (C23)

Factors related to the project manager and team members

middotProject manager Lack of commitment (C12)

middotProject manager Lack of communication skills (C13)

middotProject manager Lack of competence (C14)

middotProject manager Lack of vision (C15)

middotProject team lack of competence (C18)

middotProject team misunderstanding related to customeruser needs (C19)

middotProject team lack of commitment (C20)

Factors related to external environment

middotIrruption of competitors (C1)

middotPolitical social economic or legal changes (C11)

middotPublic opinion opposition to Project (C21)

middotUnexpected events with no effective response possible (C24)

middotUnrealistic customer expectations (C25)

FAILURE

Figure 1 Failure causes grouped by category The factors are derived from existing bibliography and previous work on the matter They havebeen classified according to Belassi and Tukel taxonomy

gathers the identified factors The factors are coded as 119862119909where 119909 is the number given to the failure cause

The factors were classified according to Belassi andTukel taxonomy [40] included in Table 1 Figure 1 depictsthe classification considered The factors are presented un-shorted because they were shuffled in order to avoid biasesin the survey The first group includes the factors related tothe organization the project belongs to (ie factors relatedto top management support) The second group includes thefactors related to the project itself and the way the projectis managed The third one comprises the factors related tothe project manager and the team members Many studiesdemonstrated the importance of selecting project managerswho possess the necessary technical and administrativeskills for successful project termination as well as theircommitmentThe competences of the teammembers are alsofound to be a critical factor Finally the factors related toexternal environment consist of factors which are external tothe organization but still have an impact on project successor failure The classification can be considered collectivelyexhaustive Belassi and Tukel state that the four groupsoffer a comprehensive set in that any factor listed in the

literature or even specific points of consideration shouldbelong to at least one group [40] However it is not easyto differentiate in which category to include each of thedifferent factors The groups are interrelated The authorsgive some indications to distinguish where to classify themmainly regarding organization and external environmentFor example if a customer is from outside the organizationheshe should be considered as an external factor (ie C25-unrealistic customer expectations) For functional projectshowever customers are usually part of the organization suchas topmanagement In such cases factors related to the clientcan be grouped under the factors related to the organization

We can find correspondences among most of the failurecauses included in our study (Table 2) and the complexityindicators appointed by Kermanshachi et al [3] For exampleC2 (which is one of the most frequent in accordance withthe rank presented in Table 7) is related to several complexityindexes as ldquolevel of project design changes derived by RFIrdquoand ldquomagnitude of change ordersrdquo C10 for example isrelated to ldquopercentage of design completed at the start ofconstructionrdquo This fact reinforces the connection betweenfailure causes and complexity factors

6 Complexity

The recipients were randomly chosen among the mem-bers of 36 project management groups from LinkedIn net-work The questionnaire was open for 3 months and 11 daysin 2011 in order to obtain a significant number of answersDuring that period of time customized emails were sentwith an invitation to answer the questionnaire to a total of3668 people A total of 619 answers were received (1688)611 of which were considered for further analysis (the restwere discarded due to consistency issues) Neither of thequestionnaires was partially filled or incomplete since allfields were marked as mandatory Previously in 2010 a pilotsurvey was conducted with the help of project managementexperts and practitioners This primary questionnaire wassent up to 45 people with a response rate of 6667 Thosefactors which reached a higher score were the ones finallyselected to configure the list included in the current studySuggestions and comments made by respondents to improvethe proposed list were also taken into account

Answers from 63 countries were received as Table 3shows

The answers have been grouped first into 13 differentgeographical zones (more details can be found in [16])according to geographical cultural historical and economiccriteria From those a total of 6 groups comprise more than85 of the respondents AMZ3 EUZ3 EUZ1 AMZ1 ASZ1and ASZ3 The other 15 of the responses were grouped in anew category called ldquoOthersrdquo Table 4 summarizes the groupsby geographical zone

Following a similar approach 53 types of projects werepresent on the answers received To simplify they have beenclassified into a total of 17 groups according to the highestlevel of the ISIC Rev 4 Classification [58] Development aidprojects were listed apart as an independent category Table 5presents the number of respondents from each project typeand the codification

Information about the size of the projects the respondentsare usually involved in was also enquired The size of theprojects is something quite subjective anddifficult to compareamong different sectors For example considering the projectbudget as an indicator of project size a project consideredsmall in construction could by contrast be consideredlarge in information technology projects In order to avoidthese biases the guideline provided by the University ofWisconsin-Madison [59] was used The results show that1195 of respondents are involved in small projects 4926in medium size projects and 3879 in large projects

Though the scope of this paper does not cover thestatistical analysis of the obtained answers here is a summaryof the main results Overall both the most and least frequentfailure causes are presented here In order to rank each one ofthe 26 failure causes a frequency index (FI) was calculated asfollows

FI = [sum611119894=1 sum4119895=1 119860 119894 lowast 119861119895]611 (1)

where119860 119894 is the number of responses choosing interval 119894119861119895 is interval factor 119895

Table 3 Countries and number of responses (ordered by numberof responses) When several countries are indicated in the samerow the number of respondents stands for the number of answersreceived from each country (ie 50 answers from Spain and 50answers from the United States)

COUNTRIES NUMBER OFRESPONDENTS

Argentina 91Spain United States 50Greece 45Chile 41India 40Brazil 30Luxembourg 27Mexico 21Uruguay 20United Arab Emirates 19Italy 18Russia 16Sweden 14South Africa 12Canada 11Germany 8Saudi Arabia The Netherlands UnitedKingdom 7

Australia Denmark Finland Kuwait 5Belgium Paraguay Slovenia 4Egypt France Israel Switzerland 3Philippines Tanzania 2Algeria Austria Bahrain Belarus BotswanaBritish Virgin Islands Cameroon ChinaColombia Cyprus Czech Republic GhanaHong Kong Hungary Indonesia IrelandJapan Jordan Nigeria Pakistan Peru PolandQatar Rwanda Singapore Surinam ThailandUganda Venezuela

1

Each interval factor is defined as presented in Table 6 Theranking is presented in Table 7

With the obtained results a cluster analysis was con-ducted to obtain patterns in the answers data set groupinga set of data with similar values The aim is to classify a set ofsimple elements into a number of groups in such a way thatthe elements in the same group are similar or related to oneanother and at the same time different or unrelated to theelements in other groups

The cluster analysis (also called Unsupervised Classi-fication Exploratory Data Analysis Clustering NumericalTaxonomy or PatternRecognition) is amultivariate statisticaltechnique whose aim is to divide a set of objects into groupsor clusters in such a way that objects in the same cluster arevery similar to each other (internal cluster cohesion) andthe objects in other groups are different (external clusterisolation) Summarizing it deals with creating data clusters

Complexity 7

Table 4 Answers grouped by geographical zone

GEOGRAPHICAL ZONESANALYZED COUNTRIES NUMBER OF

RESPONDENTSAMZ3 Argentina Brazil Chile Colombia Paraguay Peru Surinam Uruguay and Venezuela 190EUZ3 Cyprus Spain Greece and Italy 114

EUZ1 Germany Austria Belgium Denmark Finland France Ireland Virgin Islands (UK)Luxembourg United Kingdom The Netherlands Sweden and Switzerland 86

AMZ1 Canada and United States 61

ASZ1 Algeria Bahrain Indonesia Jordan Pakistan Qatar Saudi Arabia Egypt United ArabEmirates and Kuwait 40

ASZ3 India 40OTHERS Rest 80

Table 5 Project types and number of responses

CODE PROJECT TYPE NUMBER OF RESPONDENTSISIC1 Information and communication 365ISIC2 Financial and insurance activities 43ISIC3 Construction 42ISIC4 Manufacturing 41ISIC5 Professional scientific and technical activities 32ISIC6 Public administration and defence 12ISIC7 Mining and quarrying 11ISIC8 Human health and social work activities 10ISIC9 Electricity gas steam and air conditioning supply 10ISIC10 Accommodation and food service activities 8ISIC11 Education 8ISIC12 Administrative and support service activities 7ISIC13 Transportation and storage 7ISIC14 Wholesale and retail trade 6ISIC15 Real estate activities 4ISIC16 Development aid 3ISIC17 Arts entertainment and recreation 2

Table 6 Intervals and their interval factor

Interval Meaning Factor (intervalarithmetic mean)

0ndash25 SeldomUnlikely to occur 012525ndash50 OccasionalLikely to occur 037550ndash75 FrequentVery likely to occur 062575ndash100 Always occurs 0875

in such a way that each group is homogeneous and differentfrom the rest For this purposemany data analysis techniquescan be used In this study the Self-organized Maps (SOM)technique has been used a specific type of neural network[60]

SOM networks are an excellent tool for exploring andanalyzing data which are especially adequate due to theirremarkable visualization properties They create a series ofprototype vectors which represent the data set and projectsuch vectors into a low dimensional network (generally

bidimensional) from the input 119889-dimensional space whichpreserves its topology and maintains it As a result thenetwork shows the distance between the different sets sothat it can be used as an adequate visualization surface todisplay different data characteristics as for instance theircluster division Summarizing SOM Clustering Networksallow input data clustering and easily visualize the resultingmultidimensional data clusters For the analysis of datacollected in the questionnaire SOM Toolbox was used withMATLAB [61]Themethodology used is a two-level approachfor partitive clustering where the data set is first projectedusing the SOM and then the SOM is clustered as describedin [62] Partitive clustering algorithms divide a data set intoa number of clusters typically by trying to minimize somecriterion or error function An example of a commonly usedpartitive algorithm is the 119896-means which minimizes errorfunction (2) where 119862 is the number of clusters and 119888119896 is thecenter of cluster 119896

119864 = 119862sum119896=1

sum119909∁119876119896

1003817100381710038171003817119909 minus 11988811989610038171003817100381710038172 (2)

8 Complexity

Table 7 Summary of results

RANK CODE FAILURE CAUSE FREQUENCY INDEX1 C3 Customerrsquos requirements inaccurate incomplete or not defined 059922 C2 Continuous or dramatic changes to initial requirements 056653 C6 Inaccurate time estimations 053294 C16 Project requirements inadequately documented 05092

5 C10 Not or badly defined specifications at the time the Project Team startsto work 04859

6 C5 Inaccurate costs estimations 048267 C8 Lack of Management support 046058 C25 Unrealistic Customerrsquos expectations 04507

9 C4 Disagreements or conflicts of interest among different departmentsinvolved 04478

10 C19 Project Teamrsquos misunderstandings related to CustomerUserrsquos wishesor needs 04417

11 C17 Project staff changes 0430612 C22 Quality checks not or badly performed 0427013 C26 Wrong number of people assigned to the project (too many too few) 0413514 C13 Project Managerrsquos lack of communication skills 0411415 C7 Inadequate management of suppliers and procurement 0382816 C18 Project Teamrsquos lack of competence 0381517 C15 Project Managerrsquos lack of vision 0374618 C14 Project Managerrsquos lack of competence 0374219 C23 Too much complex or new technology 0357420 C20 Project Teamrsquos lack of commitment 0353721 C24 Unexpected events with no effective response possible 0329222 C9 Lack of previous identification of relevant rules and legislation 0319423 C12 Project Managerrsquos lack of commitment 0300924 C11 Political social economic or legal changes 0266225 C21 Public opinion opposition to project 0242826 C1 Competitors 02261

To select the best one among different partitionings eachof these can be evaluated using some kind of validity indexSeveral indices have been proposed [63 64] In our work weused the Davies-Bouldin index [65] which has been provento be suitable for evaluation of 119896-means partitioning Thisindex is a function of the ratio of the sum of within-clusterscatter to between-cluster separation According to Davies-Bouldin validity index the best clustering minimizes (3)which uses 119878119888 for within-cluster distance (119878(119876119896)) and 119889119888119890 forbetween-cluster distance (119889(119876119896 119876119897))

1119862119862sum119896=1

max119897 =119896

119878119888 (119876119896) + 119878119888 (119876119897)119889119888119890 (119876119896 119876119897) (3)

The approach used in this paper (clustering the SOMrather than clustering the data directly) is depicted inFigure 2 First a large set of prototypes (much larger thanthe expected number of clusters) is formed using SOM Theprototypes can be interpreted as ldquoprotoclustersrdquo [62] whichare in the next step combined to form the actual clusters Eachdata vector of the original data set belongs to the same clusteras its nearest prototype

The SOM consists of a regular usually two-dimensionalgrid of map units Each unit 119894 is represented by a prototypevector 119898119894 = [1198981198941 119898119894119889] where 119889 is input vector dimen-sion The units are connected to adjacent ones by a neigh-borhood relation The number of map units which typicallyvaries from a few dozen up to several thousand determinesthe accuracy and generalization capability of the SOM Dur-ing training the SOM forms an elastic net that folds onto theldquocloudrdquo formed by the input data Data points lying near eachother in the input space are mapped onto nearby map unitsThus the SOM can be interpreted as a topology preservingmapping from input space onto 2D grid of map units

The SOM is trained iteratively At each training step asample vector 119909 is randomly chosen from the input data setDistances between 119909 and all the prototype vectors are com-putedThe best-matching unit (BMU) which is denoted hereby 119887 is the map unit with prototype closest to 1199091003817100381710038171003817119909 minus 1198981198871003817100381710038171003817 = min

1198941003817100381710038171003817119909 minus 1198981198941003817100381710038171003817 (4)

Next the prototype vectors are updatedThe BMU and itstopological neighbors are moved closer to the input vector in

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

6 Complexity

The recipients were randomly chosen among the mem-bers of 36 project management groups from LinkedIn net-work The questionnaire was open for 3 months and 11 daysin 2011 in order to obtain a significant number of answersDuring that period of time customized emails were sentwith an invitation to answer the questionnaire to a total of3668 people A total of 619 answers were received (1688)611 of which were considered for further analysis (the restwere discarded due to consistency issues) Neither of thequestionnaires was partially filled or incomplete since allfields were marked as mandatory Previously in 2010 a pilotsurvey was conducted with the help of project managementexperts and practitioners This primary questionnaire wassent up to 45 people with a response rate of 6667 Thosefactors which reached a higher score were the ones finallyselected to configure the list included in the current studySuggestions and comments made by respondents to improvethe proposed list were also taken into account

Answers from 63 countries were received as Table 3shows

The answers have been grouped first into 13 differentgeographical zones (more details can be found in [16])according to geographical cultural historical and economiccriteria From those a total of 6 groups comprise more than85 of the respondents AMZ3 EUZ3 EUZ1 AMZ1 ASZ1and ASZ3 The other 15 of the responses were grouped in anew category called ldquoOthersrdquo Table 4 summarizes the groupsby geographical zone

Following a similar approach 53 types of projects werepresent on the answers received To simplify they have beenclassified into a total of 17 groups according to the highestlevel of the ISIC Rev 4 Classification [58] Development aidprojects were listed apart as an independent category Table 5presents the number of respondents from each project typeand the codification

Information about the size of the projects the respondentsare usually involved in was also enquired The size of theprojects is something quite subjective anddifficult to compareamong different sectors For example considering the projectbudget as an indicator of project size a project consideredsmall in construction could by contrast be consideredlarge in information technology projects In order to avoidthese biases the guideline provided by the University ofWisconsin-Madison [59] was used The results show that1195 of respondents are involved in small projects 4926in medium size projects and 3879 in large projects

Though the scope of this paper does not cover thestatistical analysis of the obtained answers here is a summaryof the main results Overall both the most and least frequentfailure causes are presented here In order to rank each one ofthe 26 failure causes a frequency index (FI) was calculated asfollows

FI = [sum611119894=1 sum4119895=1 119860 119894 lowast 119861119895]611 (1)

where119860 119894 is the number of responses choosing interval 119894119861119895 is interval factor 119895

Table 3 Countries and number of responses (ordered by numberof responses) When several countries are indicated in the samerow the number of respondents stands for the number of answersreceived from each country (ie 50 answers from Spain and 50answers from the United States)

COUNTRIES NUMBER OFRESPONDENTS

Argentina 91Spain United States 50Greece 45Chile 41India 40Brazil 30Luxembourg 27Mexico 21Uruguay 20United Arab Emirates 19Italy 18Russia 16Sweden 14South Africa 12Canada 11Germany 8Saudi Arabia The Netherlands UnitedKingdom 7

Australia Denmark Finland Kuwait 5Belgium Paraguay Slovenia 4Egypt France Israel Switzerland 3Philippines Tanzania 2Algeria Austria Bahrain Belarus BotswanaBritish Virgin Islands Cameroon ChinaColombia Cyprus Czech Republic GhanaHong Kong Hungary Indonesia IrelandJapan Jordan Nigeria Pakistan Peru PolandQatar Rwanda Singapore Surinam ThailandUganda Venezuela

1

Each interval factor is defined as presented in Table 6 Theranking is presented in Table 7

With the obtained results a cluster analysis was con-ducted to obtain patterns in the answers data set groupinga set of data with similar values The aim is to classify a set ofsimple elements into a number of groups in such a way thatthe elements in the same group are similar or related to oneanother and at the same time different or unrelated to theelements in other groups

The cluster analysis (also called Unsupervised Classi-fication Exploratory Data Analysis Clustering NumericalTaxonomy or PatternRecognition) is amultivariate statisticaltechnique whose aim is to divide a set of objects into groupsor clusters in such a way that objects in the same cluster arevery similar to each other (internal cluster cohesion) andthe objects in other groups are different (external clusterisolation) Summarizing it deals with creating data clusters

Complexity 7

Table 4 Answers grouped by geographical zone

GEOGRAPHICAL ZONESANALYZED COUNTRIES NUMBER OF

RESPONDENTSAMZ3 Argentina Brazil Chile Colombia Paraguay Peru Surinam Uruguay and Venezuela 190EUZ3 Cyprus Spain Greece and Italy 114

EUZ1 Germany Austria Belgium Denmark Finland France Ireland Virgin Islands (UK)Luxembourg United Kingdom The Netherlands Sweden and Switzerland 86

AMZ1 Canada and United States 61

ASZ1 Algeria Bahrain Indonesia Jordan Pakistan Qatar Saudi Arabia Egypt United ArabEmirates and Kuwait 40

ASZ3 India 40OTHERS Rest 80

Table 5 Project types and number of responses

CODE PROJECT TYPE NUMBER OF RESPONDENTSISIC1 Information and communication 365ISIC2 Financial and insurance activities 43ISIC3 Construction 42ISIC4 Manufacturing 41ISIC5 Professional scientific and technical activities 32ISIC6 Public administration and defence 12ISIC7 Mining and quarrying 11ISIC8 Human health and social work activities 10ISIC9 Electricity gas steam and air conditioning supply 10ISIC10 Accommodation and food service activities 8ISIC11 Education 8ISIC12 Administrative and support service activities 7ISIC13 Transportation and storage 7ISIC14 Wholesale and retail trade 6ISIC15 Real estate activities 4ISIC16 Development aid 3ISIC17 Arts entertainment and recreation 2

Table 6 Intervals and their interval factor

Interval Meaning Factor (intervalarithmetic mean)

0ndash25 SeldomUnlikely to occur 012525ndash50 OccasionalLikely to occur 037550ndash75 FrequentVery likely to occur 062575ndash100 Always occurs 0875

in such a way that each group is homogeneous and differentfrom the rest For this purposemany data analysis techniquescan be used In this study the Self-organized Maps (SOM)technique has been used a specific type of neural network[60]

SOM networks are an excellent tool for exploring andanalyzing data which are especially adequate due to theirremarkable visualization properties They create a series ofprototype vectors which represent the data set and projectsuch vectors into a low dimensional network (generally

bidimensional) from the input 119889-dimensional space whichpreserves its topology and maintains it As a result thenetwork shows the distance between the different sets sothat it can be used as an adequate visualization surface todisplay different data characteristics as for instance theircluster division Summarizing SOM Clustering Networksallow input data clustering and easily visualize the resultingmultidimensional data clusters For the analysis of datacollected in the questionnaire SOM Toolbox was used withMATLAB [61]Themethodology used is a two-level approachfor partitive clustering where the data set is first projectedusing the SOM and then the SOM is clustered as describedin [62] Partitive clustering algorithms divide a data set intoa number of clusters typically by trying to minimize somecriterion or error function An example of a commonly usedpartitive algorithm is the 119896-means which minimizes errorfunction (2) where 119862 is the number of clusters and 119888119896 is thecenter of cluster 119896

119864 = 119862sum119896=1

sum119909∁119876119896

1003817100381710038171003817119909 minus 11988811989610038171003817100381710038172 (2)

8 Complexity

Table 7 Summary of results

RANK CODE FAILURE CAUSE FREQUENCY INDEX1 C3 Customerrsquos requirements inaccurate incomplete or not defined 059922 C2 Continuous or dramatic changes to initial requirements 056653 C6 Inaccurate time estimations 053294 C16 Project requirements inadequately documented 05092

5 C10 Not or badly defined specifications at the time the Project Team startsto work 04859

6 C5 Inaccurate costs estimations 048267 C8 Lack of Management support 046058 C25 Unrealistic Customerrsquos expectations 04507

9 C4 Disagreements or conflicts of interest among different departmentsinvolved 04478

10 C19 Project Teamrsquos misunderstandings related to CustomerUserrsquos wishesor needs 04417

11 C17 Project staff changes 0430612 C22 Quality checks not or badly performed 0427013 C26 Wrong number of people assigned to the project (too many too few) 0413514 C13 Project Managerrsquos lack of communication skills 0411415 C7 Inadequate management of suppliers and procurement 0382816 C18 Project Teamrsquos lack of competence 0381517 C15 Project Managerrsquos lack of vision 0374618 C14 Project Managerrsquos lack of competence 0374219 C23 Too much complex or new technology 0357420 C20 Project Teamrsquos lack of commitment 0353721 C24 Unexpected events with no effective response possible 0329222 C9 Lack of previous identification of relevant rules and legislation 0319423 C12 Project Managerrsquos lack of commitment 0300924 C11 Political social economic or legal changes 0266225 C21 Public opinion opposition to project 0242826 C1 Competitors 02261

To select the best one among different partitionings eachof these can be evaluated using some kind of validity indexSeveral indices have been proposed [63 64] In our work weused the Davies-Bouldin index [65] which has been provento be suitable for evaluation of 119896-means partitioning Thisindex is a function of the ratio of the sum of within-clusterscatter to between-cluster separation According to Davies-Bouldin validity index the best clustering minimizes (3)which uses 119878119888 for within-cluster distance (119878(119876119896)) and 119889119888119890 forbetween-cluster distance (119889(119876119896 119876119897))

1119862119862sum119896=1

max119897 =119896

119878119888 (119876119896) + 119878119888 (119876119897)119889119888119890 (119876119896 119876119897) (3)

The approach used in this paper (clustering the SOMrather than clustering the data directly) is depicted inFigure 2 First a large set of prototypes (much larger thanthe expected number of clusters) is formed using SOM Theprototypes can be interpreted as ldquoprotoclustersrdquo [62] whichare in the next step combined to form the actual clusters Eachdata vector of the original data set belongs to the same clusteras its nearest prototype

The SOM consists of a regular usually two-dimensionalgrid of map units Each unit 119894 is represented by a prototypevector 119898119894 = [1198981198941 119898119894119889] where 119889 is input vector dimen-sion The units are connected to adjacent ones by a neigh-borhood relation The number of map units which typicallyvaries from a few dozen up to several thousand determinesthe accuracy and generalization capability of the SOM Dur-ing training the SOM forms an elastic net that folds onto theldquocloudrdquo formed by the input data Data points lying near eachother in the input space are mapped onto nearby map unitsThus the SOM can be interpreted as a topology preservingmapping from input space onto 2D grid of map units

The SOM is trained iteratively At each training step asample vector 119909 is randomly chosen from the input data setDistances between 119909 and all the prototype vectors are com-putedThe best-matching unit (BMU) which is denoted hereby 119887 is the map unit with prototype closest to 1199091003817100381710038171003817119909 minus 1198981198871003817100381710038171003817 = min

1198941003817100381710038171003817119909 minus 1198981198941003817100381710038171003817 (4)

Next the prototype vectors are updatedThe BMU and itstopological neighbors are moved closer to the input vector in

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 7

Table 4 Answers grouped by geographical zone

GEOGRAPHICAL ZONESANALYZED COUNTRIES NUMBER OF

RESPONDENTSAMZ3 Argentina Brazil Chile Colombia Paraguay Peru Surinam Uruguay and Venezuela 190EUZ3 Cyprus Spain Greece and Italy 114

EUZ1 Germany Austria Belgium Denmark Finland France Ireland Virgin Islands (UK)Luxembourg United Kingdom The Netherlands Sweden and Switzerland 86

AMZ1 Canada and United States 61

ASZ1 Algeria Bahrain Indonesia Jordan Pakistan Qatar Saudi Arabia Egypt United ArabEmirates and Kuwait 40

ASZ3 India 40OTHERS Rest 80

Table 5 Project types and number of responses

CODE PROJECT TYPE NUMBER OF RESPONDENTSISIC1 Information and communication 365ISIC2 Financial and insurance activities 43ISIC3 Construction 42ISIC4 Manufacturing 41ISIC5 Professional scientific and technical activities 32ISIC6 Public administration and defence 12ISIC7 Mining and quarrying 11ISIC8 Human health and social work activities 10ISIC9 Electricity gas steam and air conditioning supply 10ISIC10 Accommodation and food service activities 8ISIC11 Education 8ISIC12 Administrative and support service activities 7ISIC13 Transportation and storage 7ISIC14 Wholesale and retail trade 6ISIC15 Real estate activities 4ISIC16 Development aid 3ISIC17 Arts entertainment and recreation 2

Table 6 Intervals and their interval factor

Interval Meaning Factor (intervalarithmetic mean)

0ndash25 SeldomUnlikely to occur 012525ndash50 OccasionalLikely to occur 037550ndash75 FrequentVery likely to occur 062575ndash100 Always occurs 0875

in such a way that each group is homogeneous and differentfrom the rest For this purposemany data analysis techniquescan be used In this study the Self-organized Maps (SOM)technique has been used a specific type of neural network[60]

SOM networks are an excellent tool for exploring andanalyzing data which are especially adequate due to theirremarkable visualization properties They create a series ofprototype vectors which represent the data set and projectsuch vectors into a low dimensional network (generally

bidimensional) from the input 119889-dimensional space whichpreserves its topology and maintains it As a result thenetwork shows the distance between the different sets sothat it can be used as an adequate visualization surface todisplay different data characteristics as for instance theircluster division Summarizing SOM Clustering Networksallow input data clustering and easily visualize the resultingmultidimensional data clusters For the analysis of datacollected in the questionnaire SOM Toolbox was used withMATLAB [61]Themethodology used is a two-level approachfor partitive clustering where the data set is first projectedusing the SOM and then the SOM is clustered as describedin [62] Partitive clustering algorithms divide a data set intoa number of clusters typically by trying to minimize somecriterion or error function An example of a commonly usedpartitive algorithm is the 119896-means which minimizes errorfunction (2) where 119862 is the number of clusters and 119888119896 is thecenter of cluster 119896

119864 = 119862sum119896=1

sum119909∁119876119896

1003817100381710038171003817119909 minus 11988811989610038171003817100381710038172 (2)

8 Complexity

Table 7 Summary of results

RANK CODE FAILURE CAUSE FREQUENCY INDEX1 C3 Customerrsquos requirements inaccurate incomplete or not defined 059922 C2 Continuous or dramatic changes to initial requirements 056653 C6 Inaccurate time estimations 053294 C16 Project requirements inadequately documented 05092

5 C10 Not or badly defined specifications at the time the Project Team startsto work 04859

6 C5 Inaccurate costs estimations 048267 C8 Lack of Management support 046058 C25 Unrealistic Customerrsquos expectations 04507

9 C4 Disagreements or conflicts of interest among different departmentsinvolved 04478

10 C19 Project Teamrsquos misunderstandings related to CustomerUserrsquos wishesor needs 04417

11 C17 Project staff changes 0430612 C22 Quality checks not or badly performed 0427013 C26 Wrong number of people assigned to the project (too many too few) 0413514 C13 Project Managerrsquos lack of communication skills 0411415 C7 Inadequate management of suppliers and procurement 0382816 C18 Project Teamrsquos lack of competence 0381517 C15 Project Managerrsquos lack of vision 0374618 C14 Project Managerrsquos lack of competence 0374219 C23 Too much complex or new technology 0357420 C20 Project Teamrsquos lack of commitment 0353721 C24 Unexpected events with no effective response possible 0329222 C9 Lack of previous identification of relevant rules and legislation 0319423 C12 Project Managerrsquos lack of commitment 0300924 C11 Political social economic or legal changes 0266225 C21 Public opinion opposition to project 0242826 C1 Competitors 02261

To select the best one among different partitionings eachof these can be evaluated using some kind of validity indexSeveral indices have been proposed [63 64] In our work weused the Davies-Bouldin index [65] which has been provento be suitable for evaluation of 119896-means partitioning Thisindex is a function of the ratio of the sum of within-clusterscatter to between-cluster separation According to Davies-Bouldin validity index the best clustering minimizes (3)which uses 119878119888 for within-cluster distance (119878(119876119896)) and 119889119888119890 forbetween-cluster distance (119889(119876119896 119876119897))

1119862119862sum119896=1

max119897 =119896

119878119888 (119876119896) + 119878119888 (119876119897)119889119888119890 (119876119896 119876119897) (3)

The approach used in this paper (clustering the SOMrather than clustering the data directly) is depicted inFigure 2 First a large set of prototypes (much larger thanthe expected number of clusters) is formed using SOM Theprototypes can be interpreted as ldquoprotoclustersrdquo [62] whichare in the next step combined to form the actual clusters Eachdata vector of the original data set belongs to the same clusteras its nearest prototype

The SOM consists of a regular usually two-dimensionalgrid of map units Each unit 119894 is represented by a prototypevector 119898119894 = [1198981198941 119898119894119889] where 119889 is input vector dimen-sion The units are connected to adjacent ones by a neigh-borhood relation The number of map units which typicallyvaries from a few dozen up to several thousand determinesthe accuracy and generalization capability of the SOM Dur-ing training the SOM forms an elastic net that folds onto theldquocloudrdquo formed by the input data Data points lying near eachother in the input space are mapped onto nearby map unitsThus the SOM can be interpreted as a topology preservingmapping from input space onto 2D grid of map units

The SOM is trained iteratively At each training step asample vector 119909 is randomly chosen from the input data setDistances between 119909 and all the prototype vectors are com-putedThe best-matching unit (BMU) which is denoted hereby 119887 is the map unit with prototype closest to 1199091003817100381710038171003817119909 minus 1198981198871003817100381710038171003817 = min

1198941003817100381710038171003817119909 minus 1198981198941003817100381710038171003817 (4)

Next the prototype vectors are updatedThe BMU and itstopological neighbors are moved closer to the input vector in

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

8 Complexity

Table 7 Summary of results

RANK CODE FAILURE CAUSE FREQUENCY INDEX1 C3 Customerrsquos requirements inaccurate incomplete or not defined 059922 C2 Continuous or dramatic changes to initial requirements 056653 C6 Inaccurate time estimations 053294 C16 Project requirements inadequately documented 05092

5 C10 Not or badly defined specifications at the time the Project Team startsto work 04859

6 C5 Inaccurate costs estimations 048267 C8 Lack of Management support 046058 C25 Unrealistic Customerrsquos expectations 04507

9 C4 Disagreements or conflicts of interest among different departmentsinvolved 04478

10 C19 Project Teamrsquos misunderstandings related to CustomerUserrsquos wishesor needs 04417

11 C17 Project staff changes 0430612 C22 Quality checks not or badly performed 0427013 C26 Wrong number of people assigned to the project (too many too few) 0413514 C13 Project Managerrsquos lack of communication skills 0411415 C7 Inadequate management of suppliers and procurement 0382816 C18 Project Teamrsquos lack of competence 0381517 C15 Project Managerrsquos lack of vision 0374618 C14 Project Managerrsquos lack of competence 0374219 C23 Too much complex or new technology 0357420 C20 Project Teamrsquos lack of commitment 0353721 C24 Unexpected events with no effective response possible 0329222 C9 Lack of previous identification of relevant rules and legislation 0319423 C12 Project Managerrsquos lack of commitment 0300924 C11 Political social economic or legal changes 0266225 C21 Public opinion opposition to project 0242826 C1 Competitors 02261

To select the best one among different partitionings eachof these can be evaluated using some kind of validity indexSeveral indices have been proposed [63 64] In our work weused the Davies-Bouldin index [65] which has been provento be suitable for evaluation of 119896-means partitioning Thisindex is a function of the ratio of the sum of within-clusterscatter to between-cluster separation According to Davies-Bouldin validity index the best clustering minimizes (3)which uses 119878119888 for within-cluster distance (119878(119876119896)) and 119889119888119890 forbetween-cluster distance (119889(119876119896 119876119897))

1119862119862sum119896=1

max119897 =119896

119878119888 (119876119896) + 119878119888 (119876119897)119889119888119890 (119876119896 119876119897) (3)

The approach used in this paper (clustering the SOMrather than clustering the data directly) is depicted inFigure 2 First a large set of prototypes (much larger thanthe expected number of clusters) is formed using SOM Theprototypes can be interpreted as ldquoprotoclustersrdquo [62] whichare in the next step combined to form the actual clusters Eachdata vector of the original data set belongs to the same clusteras its nearest prototype

The SOM consists of a regular usually two-dimensionalgrid of map units Each unit 119894 is represented by a prototypevector 119898119894 = [1198981198941 119898119894119889] where 119889 is input vector dimen-sion The units are connected to adjacent ones by a neigh-borhood relation The number of map units which typicallyvaries from a few dozen up to several thousand determinesthe accuracy and generalization capability of the SOM Dur-ing training the SOM forms an elastic net that folds onto theldquocloudrdquo formed by the input data Data points lying near eachother in the input space are mapped onto nearby map unitsThus the SOM can be interpreted as a topology preservingmapping from input space onto 2D grid of map units

The SOM is trained iteratively At each training step asample vector 119909 is randomly chosen from the input data setDistances between 119909 and all the prototype vectors are com-putedThe best-matching unit (BMU) which is denoted hereby 119887 is the map unit with prototype closest to 1199091003817100381710038171003817119909 minus 1198981198871003817100381710038171003817 = min

1198941003817100381710038171003817119909 minus 1198981198941003817100381710038171003817 (4)

Next the prototype vectors are updatedThe BMU and itstopological neighbors are moved closer to the input vector in

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 9

N samples M Prototypes C Clusters

Figure 2 Clustering process

the input space The update rule for the prototype vector ofunit 119894 is

119898119894 (119905 + 1) = 119898119894 (119905) + 120572 (119905) ℎ119887119894 (119905) [119909 minus 119898119894 (119905)] (5)

where

119905 is time120572(119905) is the adaptation coefficientℎ119887119894(119905) is neighborhood kernel centered on the winnerunit

ℎ119887119894 (119905) = exp(minus1003817100381710038171003817119903119887 minus 1199031198941003817100381710038171003817221205902 (119905) ) (6)

where 119903119887 and 119903119894 are positions of neurons 119887 and 119894 on the SOMgrid Both 120572(119905) and 120590(119905) decrease monotonically with time

To perform the analysis a file with the following 30 inputvariables was prepared (the columns of the data matrix arethe variables and each row is a sample)

(i) Project size (this categorical variable has beenencoded as follows small = 3 medium = 32 = 9 large= 33 = 27)

(ii) 26 failure causes (Table 2)(iii) Answer ID (encoded from P1 to P611)This variable is

included only for tracing purpose during the valida-tion stage not for training

(iv) Country (the countries were grouped into 13 geo-graphical areas taking into account geographicaleconomic historical and cultural criteria)

(v) Type of project (they are encoded into 17 activitiesderived from the ISICCIIU Rev 4 codes)

For the training only the scores of the 26 failure causesand the project size were used The other variables of thedata set (country and type of project) were used just onlyin analyzing the results of the clustering The training hasbeen performed trying different grid dimensions but alwaysconsidering hexagonal topology (6 adjacent neighbors) and

trying different numbers of clusters One of the most signifi-cant pieces of information provided by this kind of analysis isprecisely the optimal number of clusters or optimal clusteringand how the samples are distributed among the clustersIn order to determine the right number of clusters theDavies-Bouldin index (DBI) was used [65] as describedabove

The resulting clusters were analyzed in order to charac-terize the perception of project manager practitioners aboutproject failure and hence complexity An analysis of whichfactors make each cluster different from the rest of the datais performed as well as what the dependencies between vari-ables are in the clusters Each cluster was featured finding theset of failure causes rated over or under the mode for theglobal survey and identifying themost remarkable differenceswith the other clusters In order to do that histograms andradar charts were used Finally we studied the populationdistribution of each cluster regarding geographical areasprojects sector and projects size The analysis is presented bymeans of contingency tables

4 Results

After several preliminary trials a 7times5 hexagonal SOM topol-ogy was chosen The SOM was trained following the methoddescribed in the former section Figure 3 shows the Unifieddistance matrix (U-matrix) and 27 component planes onefor each variable included in the training The U-matrix isa representation where the Euclidean distance between thecodebook vectors of neighboring neurons are depicted in acolor schema image In this case high values are associatedwith red colors and low values with blue colors This imageis used to visualize the data in a high-dimensional spaceusing a 2D image High values in the U-matrix represent afrontier region between clusters and low values represent ahigh degree of similarity among neurons on that region Eachcomponent plane shows the values of one variable in eachmap unit Through these component planes we can realizeemerging patterns of data distribution on SOMrsquos grid anddetect correlations among variables and the contribution ofeach one to the SOM differentiation only viewing the coloredpattern for each component plane [66]

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

10 Complexity

U-matrix

051152

Size

d

14

16

18C1

d

02022024026028

C2

d 04050607

C3

d 04050607

C4

d 03

04

05

C5

d 03040506

C6

d

04050607

C7

d

03

04

05C8

d

04

05

06

C9

d 02

03

04

05C10

d 03040506

C11

d

0202503035

C12

d 02

04

06C13

d 02

04

06C14

d 02

04

06C15

d 02

04

06C16

d

040506

C17

d 03

04

05

C18

d

030405

C19

d 03040506

C20

d 02

04

06C21

d

02

03

04C22

d 03

04

05

C23

d

03

04

05C24

d 0250303504045

C25

d 03040506

C26

d

03040506

Figure 3 SOM U-matrix and component plane matrices using a 7 times 5 hexagonal SOM

C2

(a)

C25

(b)

C9

(c)

Figure 4 Prototypes of the three patterns found in the component planes

Several similarities among the component planes of somefactors can be found in Figure 3 Three different patterns canbe clearly distinguished

(i) Pattern I C2 C3 C5 C6 C10 C16 and C17(Figure 4(a))

(ii) Pattern II C25 and C26 (Figure 4(b))(iii) Pattern III C9 C12 C13 C14 C15 C18 C20 and C21

(Figure 4(c))An example of each component plane has been depicted

in Figure 4 to illustrate each of the patterns As introduced

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 11

Table 8 Summary of patterns interpretation

Pattern number Focus attention Interpretation

Pattern I How the project is managed

This pattern of respondents links project complexity mainly with thecharacteristics of the project and how it is managed the requirements areincomplete or inaccurate with continuous changes the specifications arebadly defined and both costs and schedule are inaccurate They consider

also that the project staff changes entail a higher complexity

Pattern II Insufficient number of resources andunrealistic customer expectations

They emphasize the influence of unrealistic customer expectations and awrong (it is supposed to be insufficient) number of people assigned to the

project It is curious that neither of them is apparently under theirresponsibility (under the hypothesis that the resources assigned to the

project are a decision made by the organization managers)

Pattern III Project manager and team members skillscompetences and commitment

They have a more personalistic view of the complexity with the role of theproject manager and team members being decisive

1 2 3 4 5 608

1

12

14Davies-Bouldinrsquos index

Figure 5 Davies-Bouldinrsquos index using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axisDavies-Bouldinrsquos index)

previously we can detect correlations among variables byjust looking at similarities in the component planes [66] Wecan infer combinations of answers from the surveyed projectmanagers In this case a combination of answering patternscan be concluded from each of the three groups So theproject managers that rate factor C2 as frequent also rate asfrequent factors C3 C5 C6 C10 C16 and C17 (Pattern I)It is also the same with the other two patterns RegardingPattern I it should be noticed that all the factors consideredbelong to the project category except factor C17 that belongsto the organization category Regarding Pattern II there isan association between the rating of unrealistic customerrsquosexpectations (C25) and a wrong number of people assignedto the project (C26) Regarding Pattern III most of the factorsbelong to the project manager and team members (noted inthe figures as PMampT) category (C12 C13 C14 C15 C18 andC20)

Table 8 summarizes the interpretation of each pattern Afocus of attention has been remarked for each one and howeach pattern could be understood Obviously interpretationhas a component of subjectivity

Next 119896-means algorithm was used to build the clustersand the Davies-Bouldin index was considered to determinethe optimum number of clusters The Davies-Bouldin indexand the sums of squared errors diagrams are displayed inFigures 5 and 6 where the 119909-axis represents the numberof clusters According to the usual work methodology ofthis type of neural networks the best possible clusterizationwill be the one that reaches a better compromise in the

1 2 3 4 5 60

50

100

150

200 SSE

Figure 6 Sum of squared errors using a hexagonal 7 times 5 SOMcalculated for each clustering (119909-axis number of clusters119910-axissums of squared errors)

minimization of both parameters In this case we have taken5 different clusters

Figure 7 depicts the clustering results taking as referencethe results of the application of the 119896-means algorithm onthe optimal cluster number according toDavies-BouldinTheSOM plot sample hits (b) represent the number of inputvectors classified by each neuron The relative number ofvectors for each neuron is shown via the hexagonrsquos size andits color represents the similarity among neurons

Pearsonrsquos chi-squared test [67] was calculated to finddependence among the variables and their distribution inthe clusters The resulting 119901 values for the contingency tablesshown in Tables 9 10 and 11 were higher than 005 soit can be concluded that neither the project size nor thegeographical zone or the project type is significant for thegroups determined Nevertheless the 119901 values obtained for119862119909 are less than 005 which implies that the variation of thevalues of 119862119909 through the clusters has statistical significanceso we proceeded to the study and categorization of clustersbased on these factors with statistical significance Bar chartswere plotted to characterize each cluster (depicted in Figures8 9 10 and 11) Factors have been grouped by their categoryaccording to Belassi and Tukel taxonomy [40] in order tofacilitate the interpretation of the information

5 Discussion of Results

It can be observed that clusters 3 and 5 are associated with thehighest rates of the factors for each category while clusters

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

12 Complexity

1

2

2

2

5

5

5

1

1

2

2

5

5

5

1

1

2

4

5

3

3

1

1

4

4

3

3

3

1

1

4

4

3

3

3

(a) (b)

Figure 7 Partitioning of the SOM codebooks with 5 clusters (a) and the SOM plot sample hits (b)

Table 9 Contingency table showing the distribution of project size among the clusters

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesSize

Small 25 14 14 6 14 73Medium 84 53 68 40 56 301Large 57 35 61 42 42 237

Samples 166 102 143 88 112 611

0

01

02

03

04

05

06

07

C4 C7 C8 C17 C26

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 8 Organization factors and clusters bar chart Factorimportance average score of each cluster plotted against the modeof the global survey

1 and 2 are the lowest In fact cluster 1 only attributeshigh importancefrequency to C3 (customerrsquos requirementsinaccurate incomplete or not defined) On the other handcluster 3 rates with high values 19 in 26 factors and cluster 5rates high values 13 in 26 So the general trend is that clusters

0

01

02

03

04

05

06

07

C2 C3 C5 C6 C9 C16 C22 C23 C10

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 9 Project factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

3 and 5 rate factors equal to or over the general mode andclusters 1 and 2 rate factors equal to or below the generalmode as represented in Figures 8 9 10 and 11 An exceptionof this can be found for C5 (inaccurate cost estimations)where cluster 2 rates higher than the mode A global rating

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 13

Table 10 Contingency table showing the distribution of geographical zone among the clusters (see Table 4 for details about the geographicalzones)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesZone

AMZ1 26 9 11 5 10 61AMZ3 43 35 48 30 34 190ASZ1 8 3 12 9 8 40ASZ3 9 5 13 7 6 40EUZ1 23 18 17 10 18 86EUZ3 38 19 22 16 19 114Others 19 13 20 11 17 80

Samples 166 102 143 88 112 611

Table 11 Contingency table showing the distribution of project type among the clusters (see Table 5 for details about the project type codes)

Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 SamplesType

ISIC1 100 59 90 49 67 365ISIC2 14 5 9 3 12 43ISIC3 12 5 10 8 7 42ISIC4 7 7 12 8 7 41ISIC5 10 9 3 6 4 32ISIC6 1 0 5 4 2 12ISIC7 1 1 4 3 2 11ISIC8 3 3 3 0 1 10ISIC9 6 1 2 1 0 10ISIC10 1 2 1 1 3 8ISIC11 2 2 1 0 3 8ISIC12 3 3 0 1 0 7ISIC13 2 2 0 1 2 7ISIC14 3 2 0 1 0 6ISIC15 0 0 1 1 2 4ISIC16 0 0 2 1 0 3ISIC17 1 1 0 0 0 2

Samples 166 102 143 88 112 611

overview of each cluster can be drawn from Figure 12 wherea radar chart was built representing the average rates of eachcluster to the factors included in each category

In a similar way from Figure 13 we can conclude howeach group of factors is rated within each cluster It canbe observed that all the factors are arranged following thesame order (project organization project manager and teammembers and external factors) with three exceptions cluster1 gives PMampT the same importance as the external factors(very low in both cases) cluster 4 gives PMampT the sameimportance as organization factors and cluster 3 is the mostremarkable because it gives the same importance to theproject the organization and the PMampT factors

An interpretation of each cluster (Table 12) can be in-ferred from the presented results especially Figure 13 Inorder to facilitate the understanding the focus of attentionof each cluster is remarked

Finally we can also find some connections between thepatterns found in the component planes (summarized in

Table 8) and the clusters The most remarkable connectionscan be found within Pattern I and Cluster 1 (both sharing ahigh importance of incomplete or inaccurate requirementsthe badly defined specifications the project staff changes andinaccurate costs estimations) and Pattern III and Cluster 3(both remark the importance of project management andteam members skills competences and commitment)

6 Conclusions

The analysis performed by clustering techniques has allowedus to conclude that the total number of answers obtained canbe grouped into 5 classes of respondents who behaved dif-ferently in analyzing project failure causes and hence projectcomplexity This result is coherent with the conclusions ofthe existing literature on the subject which claims thatthere are no unique criteria and subjectivity is an inherentcharacteristic of these assessments The behavior of eachcluster can be understood by means of the bar charts shown

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

14 Complexity

Table 12 Summary of clusters interpretation

Cluster number Focus attention Description

Cluster 1Project factors specially customerrsquosrequirements inaccurate incompleteor not defined

Cluster 1 is featured as the pattern of respondents that gives very low influenceover the projects failure to the factors included in the questionnaire They justonly attribute some influence to the project and the organization categoriesThe most remarked factors are (highly rated factors that are equal or above thegeneral mode)(i) C4 Disagreement of conflict of interest among departments(ii) C17 Project staff changes(iii) C3 Customerrsquos requirements inaccurate incomplete or not defined(iv) C5 Inaccurate cost estimations(v) C10 Badly defined specifications

Cluster 2Both the project and the organizationremarking the influence of inaccuratecost estimations

Cluster 2 attributes the highest influence to the project category followed bythe factors related to the organization and to a lesser extent to the projectmanager and team members and external categoriesAll the factors related to the organization are scored with the same importanceas the general modeRegarding the project category all the factors are rated high or mediumexcepting C9 (lack of previous identification of legislation) The stress isspecially put in C5 (inaccurate cost estimations)

Cluster 3Project organization and projectmanager and team members factorsare all considered at the highest value

Cluster 3 is featured as the pattern of respondents that consider the projectmanager and team members factors as the highest value at the same level asthe project and the organization factors

Cluster 4

Project factors mainly butorganization and project manager andteam members factors are also veryimportant

Cluster 4 is characterized because it associates a medium influence to projectorganization and project manager and team member almost equally All thefactors are rated equal to the mode except three(i) C6 Inaccurate time estimations (rated below the mode)(ii) C9 Lack of previous identification of legislation (rated above the mode)(iii) C16 Project requirements deficiently documented (rated below the mode)

Cluster 5 Project and organization

Cluster 5 associates a very high influence on the projects failure to thecategories of project and organization The main difference with cluster 3 isthat cluster 5 attributes less importance to the project manager and teammembers category

0

01

02

03

04

05

06

07

C12 C13 C14 C15 C18 C19 C20

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 10 Project manager and team members factors bar chartFactor importance average score of each cluster plotted against themode of the global survey

0

01

02

03

04

05

06

07

C1 C11 C21 C24 C25

Cluster1Cluster2Cluster3

Cluster4Cluster5Mode

Figure 11 External factors and clusters bar chart Factor importanceaverage score of each cluster plotted against the mode of the globalsurvey

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 15

Project

PMampT

Organization

External

Cluster3Cluster5Cluster2

Cluster4Cluster1

Figure 12 Radar chart representing the average rate for the factorsof each category

Cluster3

Cluster5

Cluster2Cluster4

Cluster1

ProjectOrganization

PMampTExternal

Figure 13 Radar chart representing the average rate of each clusterin each category

Representing the average of each cluster for each variableand comparing it to the global mode have proven to bemeaningful in characterizing each cluster

Regarding the clusters identified and based on the dis-tribution of samples and the characterization of each onewe can conclude that they are representative in the sensethat each one has its own differential set of features It isremarkable that neither the project size nor the geographicalarea or the project type is significant considering the clustersso it can be concluded that the answers given are not specificto a particular country or typesize of project

The prevalent order of importance for the factor cate-gories is project organization project manager and teammembers and external factors The most remarkable excep-tion can be found in cluster 3 which attributes the highestinfluence also to the project manager and team memberscategory The influence of external factors is in all the casesvery low so it can be remarked that the project managersattribute the complexity to the inner features of the projectsand the management conditions

On the other hand the analysis of the component planeshas revealed that there are three clear patterns The first one(Pattern I) establishes a close association with the factorsincluded in the project category as factors of failure andhence complexity However the third pattern (Pattern III)focuses on the factors associated with the project managerand the team members This is a remarkable fact that showsat least two schools of thought the one considering thefactors inherent to the project and the one that attributesthe complexity to the skills and competences of the projectmanagers and the team members The second pattern pointsout a relation between factors C25 and C26 which belong totwo different categoriesThe analysis of the component planesis independent of the cluster analysis although some similar-ities can be found between these patterns and the clustersThemost remarkable are the similarities of Pattern I and Cluster1 and Pattern III and Cluster 3

Finally this study provides amultidimensional analysis ofthe complexity in projects Some significative combinationsof factors have been found A limitation of this study is thatonly the perception of project managers is considered Thestudy may be extended considering other project stakehold-ers Other limitations are that more than 67 of the answerscome from only 10 countries (Argentina Spain UnitedStates Greece Chile India Brazil LuxembourgMexico andUruguay) and there are a number of countries with a verylimited number of answers In addition more than 59 ofthe answers are related to IT projects Taking into account theprecedent limitations detected in the study results are likelyto be biased because of country and type of project responsepercentages and therefore they cannot be generalized Theresults have been exposed per geographical zone and perproject type to avoid the referred limitation

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work has been subsidized through the Plan of ScienceTechnology and Innovation of the Principality of Asturias(Ref FC-15-GRUPIN14-132)

References

[1] B Dao S Kermanshachi J Shane S Anderson and E HareldquoIdentifying andMeasuring Project Complexityrdquo in Proceedingsof the International Conference on Sustainable Design Engineer-ing andConstruction ICSDEC 2016 pp 476ndash482 usaMay 2016

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

16 Complexity

[2] K I Gidado ldquoProject complexity The focal point of con-struction production planningrdquo Construction Management andEconomics vol 14 no 3 pp 213ndash225 1996

[3] S Kermanshachi B Dao J Shane and S Anderson ldquoAnEmpirical Study into Identifying Project Complexity Manage-ment Strategiesrdquo in Proceedings of the International Conferenceon Sustainable Design Engineering and Construction ICSDEC2016 pp 603ndash610 usa May 2016

[4] L-A Vidal and F Marle ldquoUnderstanding project complexityimplications on project managementrdquo Kybernetes vol 37 no 8pp 1094ndash1110 2008

[5] K Remington R Zolin and R Turner ldquoA model of projectcomplexity distinguishing dimensions of complexity fromseverityrdquo in Proceedings of the 9th International Research Net-work of Project Management Conference 2009

[6] B N Baker D C Murphy D Fisher D I Cleland and W RKing ldquoProject management handbookrdquo in Project ManagementHandbookNewYork pp 902ndash919VanNostrandReinhold 1988

[7] D K Peterson C Kim J H Kim and T Tamura ldquoThe percep-tions of information systems designers from the United StatesJapan and Korea on success and failure factorsrdquo InternationalJournal of Information Management vol 22 no 6 pp 421ndash4392002

[8] T Huysegoms M Snoeck G Dedene A Goderis and FStumpe ldquoA case study on variability management in softwareproduct lines Identifying why real-life projects failrdquo Interna-tional Journal of Information Systems and Project Managementvol 1 no 1 pp 37ndash48 2013

[9] H Ernst ldquoSuccess factors of new product development Areview of the empirical literaturerdquo International Journal ofMan-agement Reviews vol 4 no 1 pp 1ndash40 2002

[10] V R Montequın S C Fernandez F O Fernandez and J VBalsera ldquoAnalysis of the Success Factors and Failure Causesin Projectsrdquo International Journal of Information TechnologyProject Management vol 7 no 1 pp 18ndash31 2016

[11] R Osei-Kyei and A P C Chan ldquoReview of studies on theCritical Success Factors for Public-Private Partnership (PPP)projects from 1990 to 2013rdquo International Journal of ProjectManagement vol 33 no 6 pp 1335ndash1346 2015

[12] J R Turner and R Muller The project managerrsquos leadershipstyle as a success factor on projects A literature review ProjectManagement Institute 2005

[13] U ShauchenkaWhyProjects Fail Uladzislau Shauchenka ISBN0987724207 2013

[14] K G Cooper J M Lyneis and B J Bryant ldquoLearning to learnfrom past to futurerdquo International Journal of Project Manage-ment vol 20 no 3 pp 213ndash219 2002

[15] T Williams ldquoAssessing and moving on from the dominantproject management discourse in the light of project overrunsrdquoIEEE Transactions on Engineering Management vol 52 no 4pp 497ndash508 2005

[16] SM Cousillas FernandezValoracion y cuantificacion de causasde fracaso y factores de exito en proyectos Universidad deOviedo 2013

[17] SM Cousillas V Rodrıguez R Concepcion and F RodrıguezldquoDentification and analysis of causes for project failure andfactors for project success in the asturian caserdquo in Proceedingsof the XIV International Congress on Project Engineering 2010

[18] D Baccarini ldquoThe concept of project complexitymdasha reviewrdquoInternational Journal of Project Management vol 14 no 4 pp201ndash204 1996

[19] M Bosch-Rekveldt Y Jongkind H Mooi H Bakker and AVerbraeck ldquoGrasping project complexity in large engineeringprojectsTheTOE (Technical Organizational andEnvironmen-tal) frameworkrdquo International Journal of Project Managementvol 29 no 6 pp 728ndash739 2011

[20] S Cicmil T Cooke-Davies L Crawford and K RichardsonExploring the complexity of projects Implications of complexitytheory for project management practice Project ManagementInstitute 2017

[21] T Cooke-Davies Aspects of complexity Managing projects in acomplex world Project Management Institute 2011

[22] C Ivory and N Alderman Can project management learn any-thing from studies of failure in complex systems ProjectManage-ment Institute 2005

[23] L-A Vidal F Marle and J-C Bocquet ldquoMeasuring projectcomplexity using the Analytic Hierarchy Processrdquo InternationalJournal of Project Management vol 29 no 6 pp 718ndash727 2011

[24] Y Lu L Luo H Wang Y Le and Q Shi ldquoMeasurement modelof project complexity for large-scale projects from task andorganization perspectiverdquo International Journal of Project Man-agement vol 33 no 3 pp 610ndash622 2015

[25] A J Shenhar O Levy andD Dvir ldquoMapping the dimensions ofproject successrdquo Project Management Journal vol 28 pp 5ndash131997

[26] D Dvir and T Lechler ldquoPlans are nothing changing plans iseverythingThe impact of changes on project successrdquo ResearchPolicy vol 33 no 1 pp 1ndash15 2004

[27] T Cooke-Davies ldquoThe lsquorealrsquo success factors on projectsrdquo Inter-national Journal of Project Management vol 20 no 3 pp 185ndash190 2002

[28] J K Pinto and J E Prescott ldquoVariations in Critical Success Fac-tors Over the Stages in the Project Life Cyclerdquo Journal ofManagement vol 14 no 1 pp 5ndash18 1988

[29] D I Cleland ldquoMeasuring Success The owners viewpointrdquo inProceedings of the 18th Annual SeminarSymposium pp 6ndash12Montreal Canada 1986

[30] L A Ika ldquoProject success as a topic in project managementjournalsrdquo Project Management Journal vol 40 no 4 pp 6ndash192009

[31] S AAssaf and SAl-Hejji ldquoCauses of delay in large constructionprojectsrdquo International Journal of Project Management vol 24no 4 pp 349ndash357 2006

[32] A P C Chan D Scott and A P L Chan ldquoFactors affectingthe success of a construction projectrdquo Journal of ConstructionEngineering and Management vol 130 no 1 pp 153ndash155 2004

[33] DWM Chan andMM Kumaraswamy ldquoA comparative studyof causes of time overruns inHongKong construction projectsrdquoInternational Journal of Project Management vol 15 no 1 pp55ndash63 1997

[34] N RMansfieldOOUgwu andTDoran ldquoCauses of delay andcost overruns in Nigerian construction projectsrdquo InternationalJournal of Project Management vol 12 no 4 pp 254ndash260 1994

[35] M Sambasivan and Y W Soon ldquoCauses and effects of delaysin Malaysian construction industryrdquo International Journal ofProject Management vol 25 no 5 pp 517ndash526 2007

[36] W Al-Ahmad K Al-Fagih K Khanfar K Alsamara S Abuleiland H Abu-Salem ldquoA taxonomy of an IT project failure rootcausesrdquo International Management Review vol 5 p 93 2005

[37] J Goedeke M Mueller and O Pankratz ldquoUncovering theCauses of Information System Project Failurerdquo 2017

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Complexity 17

[38] L J May ldquoMajor causes of software project failuresrdquo CrossTalkvol 11 pp 9ndash12 1998

[39] P Patanakul ldquoManaging large-scale ISIT projects in the publicsector Problems and causes leading to poor performancerdquoTheJournal of High Technology Management Research vol 25 no 1pp 21ndash35 2014

[40] W Belassi and O I Tukel ldquoA new framework for determiningcritical successfailure factors in projectsrdquo International Journalof Project Management vol 14 no 3 pp 141ndash151 1996

[41] C S Lim and M Z Mohamed ldquoCriteria of project successAn exploratory re-examinationrdquo International Journal of ProjectManagement vol 17 no 4 pp 243ndash248 1999

[42] T Richardson ldquoProject management pitfallsrdquo Business Commu-nications Review vol 25 p 49 1995

[43] I King ldquoThe road to continuous improvement BPR and ProjectManagementrdquo IIE Solutions vol 28 pp 22ndash28 1996

[44] G P Prabhakar ldquoWhat is project success a literature reviewrdquoInternational Journal of Business and Management vol 3 no 9pp 71ndash79 2009

[45] S M Qureshi and C Kang ldquoAnalysing the organizational fac-tors of project complexity using structural equationmodellingrdquoInternational Journal of Project Management vol 33 no 1 pp165ndash176 2015

[46] S Shafiei-Monfared and K Jenab ldquoA novel approach for com-plexitymeasure analysis in design projectsrdquo Journal of Engineer-ing Design vol 23 no 3 pp 185ndash194 2012

[47] A P C Chan D C K Ho and C M Tam ldquoDesign and buildproject success factors multivariate analysisrdquo Journal of Con-struction Engineering and Management vol 127 no 2 pp 93ndash100 2001

[48] A J Shenhar A Tishler D Dvir S Lipovetsky and T LechlerldquoRefining the search for project success factors A multivariatetypological approachrdquo RampDManagement vol 32 no 2 pp 111ndash126 2002

[49] A Tishler D Dvir A Shenhar and S Lipovetsky ldquoIdentifyingcritical success factors in defense development projects A mul-tivariate analysisrdquo Technological Forecasting amp Social Changevol 51 no 2 pp 151ndash171 1996

[50] F Marle and L-A Vidal ldquoProject risk management processesImproving coordination using a clustering approachrdquo Researchin Engineering Design vol 22 no 3 pp 189ndash206 2011

[51] FMarle L-A Vidal and J-C Bocquet ldquoInteractions-based riskclustering methodologies and algorithms for complex projectmanagementrdquo International Journal of Production Economicsvol 142 no 2 pp 225ndash234 2013

[52] L Yang C Huang and K Wu ldquoThe association among projectmanagerrsquos leadership style teamwork and project successrdquoInternational Journal of Project Management vol 29 no 3 pp258ndash267 2011

[53] T Kohonen ldquoSelf-organized formation of topologically correctfeature mapsrdquo Biological Cybernetics vol 43 no 1 pp 59ndash691982

[54] T Kohonen ldquoLearning vector quantizationrdquo in Self-OrganizingMaps pp 175ndash189 Springer 1995

[55] J V Balsera V R Montequin F O Fernandez and C AGonzalez-Fanjul Data Mining Applied to the Improvement ofProject Management Data Mining Applications in Engineeringand Medicine InTech 2012

[56] S G MacDonell ldquoVisualization and analysis of software engi-neering data using self-organizing mapsrdquo in Proceedings of the2005 International Symposium on Empirical Software Engineer-ing ISESE 2005 pp 115ndash124 aus November 2005

[57] A Naiem M El-Beltagy and S Seif ldquoProject portfolio explo-ration and visualization using self-organizing mapsrdquo in Pro-ceedings of the 10th International Conference on Informatics andSystems INFOS 2016 pp 222ndash227 egy May 2016

[58] UNS International Standard Industrial Classification of allEconomic Activities UnitedNations StatisticalOfficeNewYorkNY USA 1990

[59] PMA - Determine Project Size nd March 2018 httpspmadoitwiscedusize factorshtml

[60] T Kohonen ldquoThe Self-Organizing Maprdquo Proceedings of theIEEE vol 78 no 9 pp 1464ndash1480 1990

[61] J Vesanto JHimberg E Alhoniemi and J Parhankangas ldquoSelf-organizing map in Matlab the SOM Toolboxrdquo in Proceedings ofthe Matlab DSP conference vol 99 pp 16-17 1999

[62] J Vesanto and E Alhoniemi ldquoClustering of the self-organizingmaprdquo IEEE Transactions on Neural Networks and LearningSystems vol 11 no 3 pp 586ndash600 2000

[63] J C Bezdek and N R Pal ldquoSome new indexes of clustervalidityrdquo IEEE Transactions on Systems Man and CyberneticsPart B Cybernetics vol 28 no 3 pp 301ndash315 1998

[64] G W Milligan and M C Cooper ldquoAn examination of proce-dures for determining the number of clusters in a data setrdquoPsychometrika vol 50 no 2 pp 159ndash179 1985

[65] D L Davies and DW Bouldin ldquoA cluster separation measurerdquoIEEE Transactions on Pattern Analysis andMachine Intelligencevol PAMI-1 no 2 pp 224ndash227 1978

[66] J Vesanto ldquoSOM-based data visualization methodsrdquo IntelligentData Analysis vol 3 no 2 pp 111ndash126 1999

[67] K X Pearson ldquoOn the criterion that a given system of devi-ations from the probable in the case of a correlated system ofvariables is such that it can be reasonably supposed to havearisen from random samplingrdquo Philosophical Magazine vol 50no 302 pp 157ndash175 2009

Hindawiwwwhindawicom Volume 2018

MathematicsJournal of

Hindawiwwwhindawicom Volume 2018

Mathematical Problems in Engineering

Applied MathematicsJournal of

Hindawiwwwhindawicom Volume 2018

Probability and StatisticsHindawiwwwhindawicom Volume 2018

Journal of

Hindawiwwwhindawicom Volume 2018

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawiwwwhindawicom Volume 2018

OptimizationJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Engineering Mathematics

International Journal of

Hindawiwwwhindawicom Volume 2018

Operations ResearchAdvances in

Journal of

Hindawiwwwhindawicom Volume 2018

Function SpacesAbstract and Applied AnalysisHindawiwwwhindawicom Volume 2018

International Journal of Mathematics and Mathematical Sciences

Hindawiwwwhindawicom Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Hindawiwwwhindawicom Volume 2018Volume 2018

Numerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisAdvances inAdvances in Discrete Dynamics in

Nature and SocietyHindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom

Dierential EquationsInternational Journal of

Volume 2018

Hindawiwwwhindawicom Volume 2018

Decision SciencesAdvances in

Hindawiwwwhindawicom Volume 2018

AnalysisInternational Journal of

Hindawiwwwhindawicom Volume 2018

Stochastic AnalysisInternational Journal of

Submit your manuscripts atwwwhindawicom