DesigningaNewOrganizationatNASA: … · 2015-07-29 · 2.1....

OrganizationScienceVol. 17, No. 2, March–April 2006, pp. 202–214issn 1047-7039 �eissn 1526-5455 �06 �1702 �0202

informs ®

doi 10.1287/orsc.1050.0166©2006 INFORMS

Designing a New Organization at NASA:An Organization Design Process Using Simulation

Timothy N. CarrollCollege of Management, Georgia Institute of Technology, 800 West Peachtree Street, N.W., Atlanta, Georgia 30308,

[email protected]

Thomas J. GormleyGormley & Associates, 6 Recodo, Irvine, California 92620, [email protected]

Vincent J. BilardoNASA Glenn Research Center, 21000 Brookpark Road, Mailstop 86-8, Cleveland, Ohio 44135, [email protected]

Richard M. BurtonFuqua School of Business, Duke University, Durham, North Carolina 27705, [email protected]

Keith L. WoodmanNASA Langley Research Center, Hampton, Virginia 23681, [email protected]

The challenge for NASA’s Systems Analysis Integrated Discipline Team (SAIDT) is to develop a new organizationdesign capable of performing complex modeling and analysis tasks, using team members at various NASA centers.

The focus is on: (1) design as a process, (2) the effect of design tools on the process as well as alternative designs,(3) the fit between the tools and their fit with the organization, (4) the effect of an ongoing agencywide transformation,and (5) implications for organizational contingency theory.

Key words : organization design; NASA; simulation; process; tools; contingency theory

1. IntroductionNASA faces the challenge of managing and coordinatingdistributed technical experts developing the next gener-ation of space exploration systems. We report on theefforts of NASA managers designing the Systems Anal-ysis Integrated Discipline Team (SAIDT), one of severalteams executing NASA’s Vision for Space Exploration(2004). The larger organization and environment con-strained the design, and the design needed to coordinateSAIDT work tasks involving personnel from geograph-ically distributed centers. The design process demon-strated new ways of organizing and conducting businessafter the shuttle Columbia disaster. Thus, not only theresulting design, but the process was pushing the agendafor change within NASA.As such, this paper is as concerned with the pro-

cess of discovery and refinement (Nystrom et al. 1976,Weick 1977, Pettigrew 1987, Greenwood and Hinings1988) as with the design product. The iterative pro-cess is also shaped and constrained by the design tools.It improves the design by developing choices. Misfitsare identified and fixed. Designers also review progressmade and problems encountered, and make midcourseadjustments.Three tools used in this study shaped the questions

considered, the data available, and the possible answers.The Design Structure Matrix (DSM) (Steward 1981)

provided a static look at interdependencies betweenorganizational units and project tasks. OrgCon™ is anexpert system program that assessed the fit betweenmultiple organizational and environmental contingen-cies for the organization as a whole (Burton and Obel2004). SimVision™ gave a “bottoms-up” view of taskflows, actors, and interdependencies to predict time, cost,and quality. These tools fit the culture of the organi-zation. NASA relies heavily on engineering and com-putational techniques, especially simulations. Simulatingthe human systems fostered initial acceptance by theinvolved project managers.This paper begins with some background on NASA,

then discusses the process followed in developing thenew organization designs. The first phase of analysisformed integrated engineering teams using the DSMapproach. The second phase relied on OrgCon to developa baseline model for the SAIDT. The final phaseexplored alternatives and assessed the robustness of thebaseline design, using both SimVision and OrgCon. Thispaper concludes by discussing implications for organi-zation theory.

2. The NASA Organization DesignChallenge

Understanding NASA’s contextual frame of reference isessential to an understanding of how the organization

202

Carroll et al.: Designing a New Organization at NASAOrganization Science 17(2), pp. 202–214, © 2006 INFORMS 203

design process unfolded (Van de Ven 1992). Like allorganizations, NASA’s internal and external contextsboth enable and constrain the process of generatingalternatives and evaluating potential designs (Pettigrew1992).The United States Vision for Space Exploration

(2004) presented NASA with an overall mission thatincluded a return to the Moon followed by human explo-ration of Mars and beyond. Mindful of the difficultyof managing complex, large-scale technical projects,NASA managers recognized that organization design,and the ability to evolve the design over a long durationprogram would be critical to success. Thus, NASA man-agers sought better methods to model and analyze thecharacteristics and performance of their units, includ-ing their structure, cost, internal and external interac-tions (with industry partners, for example), and overalleffectiveness.

2.1. NASA’s Exploration Program—ConstellationSystems

The SAIDT is one component of Project Constellation.Considered in its entirety, Constellation Systems refersto the complete set of systems required for human andhuman/robotic exploration activities on the Moon, Mars,and beyond.The components of Constellation are being developed

in a modular fashion. The large number and complex-ity of the design options for the components themselvesis compounded by the additional need for these com-ponents to work together. For example, it is criticallyimportant that the Crew Exploration Vehicle (CEV), itslaunch vehicle, the lunar lander, and other equipment,which share functional interfaces with the CEV, all sharea robust, interoperable software architecture. This archi-tecture must be “open” in the sense that it be structuredto gracefully accept the inevitable upgrades in softwareapplications that will be produced over a 20-year pro-gram life cycle.The integration of the engineering systems is paral-

leled by the required integration of the organizationalunits building the components. The staff responsiblefor developing software applications reside in multi-ple, geographically distributed organizations that mustshare similarly open interfaces. The boundaries that mustbe spanned across these interfaces include not onlygeographical boundaries, but also government-industryboundaries and multiple, often-competing contractorswithin industry.The resulting organizational design challenge is to

create a high-performing organization characterized bywide-open communication across the many interfaces,strong motivation to cooperate in spite of competingprofit motives, and strong goal orientation to ensureastronaut safety and mission success.

2.2. The Ongoing NASA TransformationEven before the Vision for Space Exploration wasannounced, NASA had begun an agencywide effort toaddress cultural and organizational deficiencies that con-tributed to the loss of 2 space shuttles and 14 astronautsover 14 years. The Columbia Accident InvestigationBoard (CAIB) report concluded that deficiencies in man-agement and organization were a main contributor to theaccident:

The organizational causes of this accident are rootedin the Space Shuttle Program’s history and culture,including � � � resource constraints, fluctuating priorities,schedule pressures � � � and lack of an agreed nationalvision for human spaceflight. Cultural traits and organi-zational practices detrimental to safety were allowed todevelop, including: reliance on past success as a sub-stitute for sound engineering practices (such as test-ing to understand why systems were not performingin accordance with requirements); organizational barri-ers that prevented effective communication of criticalsafety information and stifled professional differencesof opinion; lack of integrated management across pro-gram elements; and the evolution of an informal chainof command and decision making processes that oper-ated outside the organization’s rules. (CAIB Report 2003,Vol. I, p. 177)

Because organizational causes contributed to the Chal-lenger and Columbia accidents, new organizationdesigns at NASA must facilitate quality decision mak-ing (Starbuck and Milliken 1988). The design must alsopromote the “One NASA” initiative, which aims to buildprocesses, tools, and capabilities for better communica-tion and collaboration among the 10 NASA field centers.NASA’s organizational structure was reorganized in

January 2004 to realign internal operations and decision-making authority with the new exploration agenda.The new structure was based on four “mission direct-orates”—Exploration Systems, Space Operations, Sci-ence, and Aeronautics Research—and a handful of head-quarters support functions. The Exploration SystemsMission Directorate is responsible for Constellation, yetthe personnel who will staff Constellation projects workin 10 field centers managed by the other 3 mission direc-torates. Thus the organization structure is a large matrixbetween projects (managed by Exploration Systems) andfunctional areas (overseen by Space Operations, Science,and Aeronautics Research).

3. Formulating the Integrated DisciplineTeams (IDTs)

3.1. NASA’s Organization Design TeamNASA established the Organization Design Team (ODT)in 2003 to research methods for designing the “humansystems,” e.g., the programs, projects, and implement-ing organizations, which are a critical component

Carroll et al.: Designing a New Organization at NASA204 Organization Science 17(2), pp. 202–214, © 2006 INFORMS

of the hardware and software systems that performNASA’s missions. Although the ODT formed prior tothe Columbia accident, the aftermath of the accidenttogether with the reforms called for in the CAIB reportprovided a sense of urgency to the ODT’s search forimproved organizational design capabilities.

3.2. Constellation Integrated Discipline TeamFormulation Study

Because the Project Constellation effort spans all ofNASA’s 10 field centers as well as headquarters, Con-stellation senior managers instructed the ODT to con-duct an organization design study that would establishan organization capable of accessing the technical staffwho possess the needed skills and disciplines at the fieldcenters. The ODT, with expert input from the 10 fieldcenters, developed a teaming structure aligned aroundtechnical disciplines called Integrated Discipline Teams(IDTs).The methodology relied heavily on interviews, con-

sensus discussions and iterative improvements usingmatrices developed with the DSM (Steward 1981). Thistool allows for graphical representation of interfaces andinformation flows between actors in an organization,hardware or software elements in a complex system,or tasks in a program or project, including tasks thatare iterative or require feedback. Although the DSMmethodology is static and lower in fidelity than othertools (such as SimVision), its generality and ease of usemeant that it was well suited to aid in the initial designof the integrated discipline teams. Because the IDTswere formed from technical personnel in the field centersworking on aspects of Constellation that shared interde-pendencies, the DSM methodology seemed appropriate.

3.3. Conducting the IDT AnalysisThe study team decided to group personnel on the basisof skills/disciplines rather than products because theConstellation Systems mission had just been created andno products had been identified. The IDTs were formedby analyzing critical skill interactions (see Figure 1)using the DSM analysis approach. After individuallyassessing the degree of interaction and meeting todiscuss those areas with high levels of disagreementbetween team members, the team clustered the highinteraction items along the diagonal of the matrix. Theclustered items represent skills with high levels of inter-action, which provided the logic for the IDTs.

3.4. IDT Formulation Study ConclusionSenior Constellation managers supported this process,and implemented the recommended 14 IDTs with onlyminor changes (see Figure 2). These teams providedConstellation Systems with access to 200 full-timeequivalent personnel across 10 field centers to initi-ate the complex Constellation Systems design process.

The ODT leader then transitioned to lead one particu-lar IDT, the SAIDT. In addition, the ODT was assignedto the SAIDT as a subteam to transition the Program/Organization Modeling and Simulation (POMS) initia-tive without loss of momentum.

4. Designing the SAIDTIn Phase 2, the design team sought to find an acceptableorganizational design for the Constellation SAIDT. TheSAIDT was chartered to perform technical studies andsystems analysis tasks in support of the design, develop-ment, test, and evaluation of Constellation Systems. Thetask for the ODT was to design a cohesive team thatwould draw on the technical expertise of personnel fromall 14 IDTs staffed from up to 9 NASA centers. Thisrequired the SAIDT chairperson and his core leadershipteam to identify the personnel requirements, estimate theresources required for the team, and scope the roles andresponsibilities of individual team members. To facili-tate the organization design process for SAIDT, the ODTneeded to develop an understanding of the internal divi-sion of work and coordination mechanisms, as well astheir likelihood of meeting project goals (time, resourcesrequired, and quality). These were design aspects at alevel of detail that the DSM analysis could not address.

4.1. SAIDT Design Study ChallengeThe first of three sets of organization design questionsrelated to the best way to draw on the expertise locatedin geographically distributed field centers. The secondset dealt with coordination mechanisms. The third setrelated to performance. How likely would the SAIDT beable to meet schedule, cost, and quality goals? In addi-tion to these questions, the SAIDT chairperson set forthrequirements for the organization. It needed to be flexi-ble, because each study task would require a team to beformed, analysis done, and results disseminated quickly(typically two to four months). And it needed to haveclear lines of responsibility, as the CAIB report (2003)had identified unrealistic goals, schedule pressures, anda lack of clear delegation of roles and responsibilities asindirect causes of the Columbia accident.

4.2. Developing the SAIDT Baseline ModelThe first concern of the ODT related to how virtualthe SAIDT should be. The first scenario, termed theCore Team Option, relied on moderately skilled person-nel colocated at only one or two NASA centers, andpermanently assigned to the SAIDT on at least an annualbasis. The advantage of easier face-to-face coordinationwould be offset by the disadvantage of using perma-nently assigned staff who might not be as technicallyskilled as others and who might only be available ona part-time basis because of other priority assignments.The study personnel would also include people who hadnot worked together in the past.


Figure1

IDTFormulationDesignStructureMatrix

•A

eros

cien

ces

and

flig

ht m

echa

nics

•C

omm

and,

con

trol

, and

com

mun

icat

ions

•C

ompu

ters

, sof

twar

e, a

utom

atio

n, a

nd r

obot

ics

•H

uman

cen

tere

d

•Po

wer

sys

tem

s

•Pr

opul

sion

and

flu

ids

•St

ruct

ural

, mec

hani

cal,

mat

eria

ls, a

nd m

anuf

actu

ring

•Pr

ogra

m s

uppo

rt a

nd b

usin

ess

oper

atio

ns

•Sa

fety

and

mis

sion

ass

uran

ce

•Sy

stem

s en

gine

erin

g an

d in

tegr

atio

n

•C

onst

ella

tion

syst

ems

anal

ysis

•C

ost e

ngin

eeri

ng

•G

roun

d in

fras

truc

ture

•O

pera

tionsD

SM te

chni

que

was

use

d to

clu

ster

ski

lls

havi

ng h

igh

degr

ee o

f int

erde

pend

ency


Figure 2 Constellation IDTs

The alternative design was termed the DistributedMatrix Team Option. Here, a number of highly skilledstaff drawn from many of the NASA field centers wouldcoordinate via frequent, virtual meetings using telecon-ferences and webcasts. The ODT wanted to considerwhether there was a limitation to the number of partic-ipating centers such that efficiency (and/or other mea-sures of organizational performance) decreases as thenumber of participants increases.During this phase of work, the ODT relied primar-

ily on OrgCon, an expert system informed by multicon-tingency theory (Baligh et al. 1996, Burton and Obel2004). In an OrgCon analysis, users input information onthe organization’s environment and organizational designchoices and OrgCon provides a top-down assessmentof the fits and misfits between contingencies such asthe organization’s strategy, structure, incentives, man-agement style, climate, and environment.The ODT believed that most NASA managers, faced

with the need to design a new project team, would relyon their past experience and intuition to develop an orga-nization design. They thought that tools such as OrgConand SimVision (described in §5) could supplement intu-ition with tools that allowed managers to evaluate alter-native organization designs and to develop insight intopotential problem areas before the project began.The process of developing the baseline OrgCon model

actually answered this question before the simulation

itself was run. That is, the process of providing inputsto the OrgCon model gave the team a different view ofthe organizational design than they had up to this point.The model development process relied on interviewingthe chairperson and deputy chairperson of the SAIDT toanswer the roughly 60 input questions for the OrgConmodel dealing with the organization’s strategy, structure,and environment (see Table 1). The managers were firstinterviewed individually, then interviewed together todevelop a consensus set of answers. This process helpedcreate a common view of the SAIDT organization designand operating challenges.After discussing these options and developing the

model inputs, the ODT decided that, in reality, the CoreTeam and Distributed Matrix Team were not equallyplausible discrete options, given the NASA context. Forinstance, the SAIDT chairperson recognized that moststudy tasks would be temporary, somewhat foreseeable,relatively high-priority efforts conducted in finite cyclesof two to four months’ duration. In all likelihood, hewould have to negotiate with functional line managersat the field centers to request staff with the requisiteskills and experience be assigned to his study tasks,particularly because the skills needed would vary fromstudy to study, and from cycle to cycle. Thus the CoreTeam Option in which the SAIDT would be populatedwith permanently assigned staff with a finite skill set


Table 1 OrgCon Inputs

Input questions Baseline model characteristics

Organizational Simple (cf. 65%) or functionalconfiguration (4) (cf. 65%) or fatrix (cf. 61%)

Organizational High (cf. 100%)complexity (7)

Centralization (10) Medium (cf. 100%)Formalization (7) Medium (cf. 100%)Incentives (1) Group-based results (cf. 100%)Size (1) Medium size (cf. 100%)Leadership and Entrepreneurial (cf. 93%)

management style (6)Organizational climate (7) Developmental (cf. 77%) or

group (cf. 76%)Technology (6) Semiroutine (cf. 100%)Environment (4) High complexity (cf. 100%),

high uncertainty (cf. 100%),medium equivocality (cf. 100%),high hostility (cf. 100%)

Strategy (5) Analyzer with innovation (cf. 72%)

Notes. cf. stands for the certainty factor, a measure of the confi-dence the OrgCon expert system has in its assessment of eachorganizational characteristic.

The number in parentheses following each input question cate-gory is the number of OrgCon input questions for that category.

was not a realistic option. In addition, a team struc-ture was needed in which a small leadership team couldhandle the daily team management, planning, and tech-nical oversight to maximize team agility and minimizebroader staff resource utilization until needed to executespecific study tasks.A more plausible option was a type of matrix where

the study task leads receive direction from the SAIDTPlanning Team managers (see Figure 3). In this case, the

Figure 3 Baseline SAIDT Organization

Analysis teamPlanning team

System engineeringand integration IDT

Other IDTs

Organizationdesign team

Study task performance

Study results synthesis

• Approximate 25 analysts from 7

supporting NASA centers

• Woodman/Langley (author)

• Gormley/SAIC (author)

• Four additional members

Analysis planningAnalytical consistency

Study review and QC

• Ames lead

• Glenn lead

• Johnson lead

• JPL lead

• Kennedy lead

• Langley lead

• Marshall lead

Organization design studies

Systems analysis IDT• Vincent Bilardo/Langley (author)

Note. QC = quality control; JPL = Jet Propulsion Laboratory; SAIC = Science Applications International Corporation.

study task leads were drawn from seven of the support-ing NASA field centers and direct their assigned studyteam members, also from the various centers. The studyteam members then answer to the study task leads forthe duration of the study only, as assigned by their lineorganization supervisors based on skills and experience.This group loosely constitutes the Analysis Team of theSAIDT, as shown in Figure 3.The baseline SAIDT was comprised of about 30–40

people, many working part time (resulting in about 20full-time equivalents). The staff were drawn from sevendifferent NASA centers, with approximately half of thestaff from the two lead SAIDT centers (Langley andJohnson). There were typically two to three line organi-zations from which Analysis Team members were drawnfrom within each center as well. The SAIDT PlanningTeam, which consisted of a core group of about 10 staff,provided direction on analysis study task definition andpriorities, and the Analysis Team members executed theindividual study tasks, with each study led by a taskmanager who was delegated responsibility from the teamchairperson to execute the study within the assignedresources and required schedule.

4.3. SAIDT Baseline Model ResultsThe ODT then modeled the baseline SAIDT orga-nization design option in OrgCon. OrgCon considerswhether any of the environmental, organizational, andstrategic contingencies are misfit with each other. TheSAIDT baseline organization design presented almost nomisfits or unbalanced situations among the contingencyfactors of management style, size, environment, tech-nology, climate, strategy, and the organization structure.The only misfit suggested was based on the organizationhaving a developmental climate (see Table 2), which is


Table 2 OrgCon Results

Case Description Strategic misfits Contingency misfits

Case 0 Baseline with direct work only None • Developmental climate �= high organizationalcomplexity

Case 1 Full baseline case None • Developmental climate �= high organizationalcomplexity

Case 2 Add personnel to critical Group climate �= analyzer with • Group climate �= high organizational complexitypath tasks innovation strategy

Case 3 Eliminated an analysis cycle None • Developmental climate �= high organizationalcomplexity

Case 4 Case 3 with high centralization None • Developmental climate �= high organizationalcomplexity

• Developmental climate �= high centralization

Case 5 Case 3 with low centralization None • Leader (lead style) �= high formalization• Leader (lead style) �= functional configuration• Leader (lead style) �=machine bureaucracy• Developmental climate �= functional

configuration• Developmental climate �= high organizational

complexity• Developmental climate �= high formalization

Case 6 SAIDT Tasks with pre-NASA • Analyzer strategy �=management • Manager (lead style) �= results-based incentivestransformation characteristics with short time horizon • Group climate �= functional organization

• Group climate �= leaders with high structurepreference for making decisions • Group climate �= high degree of formalization

• Group climate �= analyzer withinnovation strategy

Note. �= denotes a contingency misfit.

misfit with a high degree of organizational complexity.OrgCon suggests that a developmental climate demandsnot to be restricted by a rigid hierarchy and job spe-cialization, so that everyone can react quickly to newsituations (Burton and Obel 2004). However, the SAIDTorganization was almost equally likely to have a groupclimate as a developmental climate. Because a groupclimate is not a misfit with high complexity, the pro-posed SAIDT organization design was predicted to befunctional and reasonable. This demonstrates that it isin alignment from an organization theory standpoint, butnot that it is the only or the ideal alternative. When theorganization design is implemented, the climate must bean ongoing concern of the management to counter therigidity of the structure and not let the job specializationhinder its operations, which will become a major part ofthe leader’s job.In summary, the second phase of this study formed

the baseline organization design for the SAIDT. Froma NASA perspective, this design was feasible and rea-sonable. From a contingency theory perspective, it pre-sented only one possible misfit, although this was notdefinitive. The next phase of this study focused on thenext tier of greater detail, working within the estab-lished macrostructure to develop the more microlevel,task-oriented design to assess the performance of themicrodesign, and to develop alternative task designswithin the overall macrostructure.

5. Refining the SAIDT OrganizationDesign

5.1. SAIDT Refinement StudyThe ODT next analyzed the baseline model usingSimVision, a system modeling, project design, and dis-crete event simulation tool (see Levitt et al. 1994, 1999for more on the development of the Virtual DesignTeam, the precursor to SimVision). In a SimVision anal-ysis, users input information on project parameters, taskflow and task features, and individual actor character-istics. The model then allows quantitative analysis oflargely qualitative phenomena to develop an organizationthat has predictable capabilities and known limits. TheODT developed a baseline organization structure, a cor-responding schedule of tasks and milestones, and a setof program risk factors. Similar to the OrgCon process,the ODT held review meetings with SAIDT membersto develop and validate the baseline microlevel SAIDTSimVision models.

5.2. Case 1: The SAIDT Baseline ModelThe SAIDT Baseline Model combined elements of afunctional, a simple, and a matrix organization. The ini-tial baseline model (Case 0) captured actors, tasks, andmilestones for the SAIDT project (see Table 3 for detailson the model settings). The SAIDT project required acyclical execution of a planned set of complementary


Table 3 SimVision Settings

Properties Case 0 Case 1 Case 2 Case 3 Case 4 Case 5

Team experience Medium Medium Medium Medium High LowCentralization Medium Medium Medium Medium High LowFormalization Medium Medium Medium Medium Low HighMatrix strength Medium Medium Medium Medium Low High

Notes. Team experience affects an individual’s information process-ing speed.

Centralization is based on where decision-making authorityresides.

Formalization affects the likelihood of formal or informal commu-nication.

Matrix strength models the “connectedness” of an organizationby setting the probability that workers will attend to exchanges ofinformation.

engineering analyses and trade studies to characterizethe capability and functionality of the Constellation Sys-tems. The Case 0 model is essentially a critical pathmethod version of the project that does not include theeffects of “hidden work” such as coordination, rework,or missed communications. The ODT personnel vali-dated the Case 0 model by comparing the SimVisionschedule and manpower data against the SAIDT ProjectSchedule. Upon validation that the Case 0 model accu-rately represented their planned project, the ODT devel-oped a more articulated and realistic baseline model(Case 1), which incorporated many more details on theactors (skill types and levels, full-time equivalents), tasks(communication and rework links between tasks), andproject parameters (such as centralization, formalization,and team experience as well as information exchangeand error probabilities).The SAIDT Case 1 Simulation predicted the impact

of rework and communication failure on several of theproject’s original performance objectives. It indicatedthat the project would likely miss its schedule comple-tion date by two months and would require more totalwork days than forecast (see Table 4). From a qualitystandpoint, however, the results were satisfactory. Therisk to quality from functional exceptions was 0.468,where 0.4 is significant but acceptable and representsproblems that do not propagate, as they do not affectother tasks. The risk to quality arising from projectexceptions was 0.251, an acceptable level of risk. Thekey result of this effort was the recognition that afteradding in realistic project details beyond the critical pathmodel, the simulation predicted that the SAIDT wasunlikely to achieve their scheduled completion date.From a macrocontingency theory perspective, the base-

line SAIDT organization performs reasonably well (seeTables 1 and 2) based on the OrgCon analysis. From aninformation processing viewpoint, however, the baselinemodel is projected to miss the schedule goals by morethan two months (see Table 2) based on the SimVisionanalysis. As the CAIB report (2003) points out, sched-ule pressures, resource constraints, and organizational

Table 4 SimVision Results

DirectProject project Total Indirect Functional Projectduration work personnel work risk risk

Case (days) (days) days (days) index index

0 394 391 5,587 165 0.000 0.0001 442 395 5,899 472 0.468 0.2512 427 373 6,254 468 0.474 0.2513 398 344 5,508 461 0.466 0.2524 410 341 5,640 591 0.321 0.2385 397 343 5,458 413 0.617 0.323

Notes. Project duration: The calendar time required for the projectto complete.

Direct project work: Time spent by team members on task work.Total personnel days: The sum of individual team members work

time.Indirect work days: Time spent on rework, coordination, and wait

time.Functional risk index: Measures the risk to quality arising from

functional exceptions—problems that affect only the task fromwhich they arise.

Project risk index: Measures the risk to quality arising from projectexceptions—problems that may have an effect on work in anothertask.

culture issues, such as barriers to effective communi-cation and a lack of a functioning chain of command,were all factors in the shuttle disasters. Thus the ODTdeveloped several alternative designs, seeking to reducecompletion time while preserving quality standards andminimizing misfits.

5.3. Case 2: Add Personnel Performing Tasks onthe Critical Path

The ODT wanted to understand the effect of adding per-sonnel to perform tasks on the critical path. Intuitively,one might expect additional resources to improve theproject completion time. However, many years of man-agerial practice in areas such as software engineeringhave shown that adding personnel to a late project may,in fact, cause the project to delay further (Brooks 1974).Adding personnel in the OrgCon model produced two

misfits, one with the strategy and the other with the highdegree of organizational complexity (see Table 2), indi-cating diminished fit between organizational, strategic,and environmental contingencies. The SimVision modelshowed mixed performance. Although Case 2 took lesstime to complete (427 days versus 442 for Case 1, thefull baseline model), the schedule improvement came atthe expense of additional work days (6,254 versus 5,899,see Table 4).

5.4. Case 3: Limit the Scope of WorkBecause additional resources did not improve on theoverall performance of the baseline case, the most obvi-ous options were to delay the schedule or reduce thework volume. As other work flows are contingent uponthe timely completion of the SAIDT work, the ODT


next studied the effect of reducing the work volume bylimiting the scope of one of the analysis cycles.While the elimination of tasks has significant effects

for the bottom-up, task-oriented SimVision analysis, itchanges nothing for the top-down, organizational con-tingency theory-based OrgCon analysis. Thus, resultsfor Case 3 were the same as the baseline model (seeTable 2). The SimVision analysis shows that reducingthe scope of work significantly improved the model(see Table 4), reducing the project duration (398 days)and total work involved (5,508 days), while maintain-ing acceptable levels of risk as measured by functional(0.466) and project (0.252) exceptions.The results of Case 3 made a strong argument to

limit the scope of work to meet schedule and budgetgoals. Again, the CAIB report on the Columbia disasterattributed the loss of the shuttle, in part, on pressures tomeet schedule goals. The ODT wanted to avoid creat-ing a similar situation by setting unrealistic schedule andbudget goals. Given that Case 3 looked like a reasonableapproach to setting up the SAIDT, the team next workedto refine this organization design.

5.5. Case 4: Case 3 with High CentralizationThe SAIDT chairperson wanted to know what level ofcentralization made the most sense. Higher centraliza-tion provides more expert direction from project man-agers who can integrate the results of the projects, anduse that information to inform their recommendations tohigher management. On the other hand, high centraliza-tion also offers the prospect of information bottlenecksand delegation by default—when lower level engineersare forced to make their own decisions after waiting toolong for direction from their managers (Kim and Burton2002).The ODT developed Case 4 as a highly centralized

organization with the same task structure as Case 3. InOrgCon, Case 4 shows the same contingency misfit asthe baseline model—the developmental climate is mis-fit with the high degree of organizational complexity(see Table 2). A developmental climate should not berestricted by a rigid hierarchy and job specialization, soeveryone can react quickly to new situations (Zammutoand Krakower 1991). The developmental climate is alsomisfit with a high degree of centralization. A devel-opmental climate requires a decentralized organizationstructure to be able to react quickly to new events andopportunities (Burton et al. 2002).By using SimVision, Case 4 increased the project

duration (to 410 days) and the total work days (to 5,640).As expected, functional exception quality improved(0.321), as subordinates were more likely to seek direc-tion from their managers when encountering an excep-tion (see Table 4). However, because the previous levelof quality was acceptable, this does not represent amajor improvement. In short, the change to a higher

level of centralization resulted in an additional contin-gency misfit, projected problems meeting the schedule,and additional resources were required. Thus, Case 4 isnot expected to be an improvement over Case 3; the bestmodel examined to this point in the study.

5.6. Case 5: Case 3 with Low CentralizationThe Case 5 analysis is the converse of Case 4. Becausethe increase in centralization diminished project teamperformance, the ODT next developed a highly decen-tralized model to determine if the opposite approachwould improve project performance.The OrgCon analysis produced a number of misfits

related to the leadership style and organizational climate(see Table 2). Unlike the “entrepreneur” leadership styleof the previous OrgCon models, the “leader” style of thedecentralized model is associated with a high preferencefor delegation and a low level of detail in decision mak-ing. This style is a misfit with elements of NASA suchas a high formalization, a functional configuration, anda machine bureaucracy as these organizational elementsare too rigid and inflexible for the style of leader. Thedevelopmental climate is a dynamic and flexible organi-zational climate based on a high level of trust, and thusis a misfit with a functional configuration, high formal-ization, and high organizational complexity, as they tendto also be rigid and inflexible (Zammuto and Krakower1991).The SimVision analysis shows that compared with

Case 3, the highly decentralized Case 5 has a nearlyidentical duration (397 for Case 5 versus 398 for Case 3)and cost (5,458 days for Case 5 versus 5,508 for Case 3),but that quality is notably worse (see Table 4). The func-tional risk in particular (0.617 for Case 5 versus 0.466for Case 3) is approaching dangerous levels (0.7 is con-sidered very high and problematic). The OrgCon andSimVision cases together demonstrate that Case 5 doesnot improve likely project performance, and thus Case 3is the best option considering project duration, cost, andquality together.

5.7. Case 6: What If SAIDT Had an OrganizationDesign Like Pre-2003 NASA?

One of the major motivations for the new NASA orga-nization design was to move away from a flawedprevious model. The next analysis demonstrates thepositive aspects of the ongoing change, i.e., that with-out the changes underway, the organization would facemany more problems. This analysis employed OrgCon,because many of the relevant changes were in areas thatOrgCon is better suited to address such as culture, lead-ership, structure, and strategy (see Table 1 for inputs tothe OrgCon model).Implementing the tasks of SAIDT with an organiza-

tion based on the old NASA resulted in a large numberof misfits (see Table 2). The CAIB report pointed out


that NASA managers felt pressured to maintain a short-term focus on schedule and budget at the expense ofquality processes. This short time horizon approach con-flicts with an analyzer with innovation strategy (Milesand Snow 1978). This strategy, which requires flexibilityto change, also conflicts with a group climate character-ized by high resistance to change. In addition, a groupclimate with a high degree of trust, low conflict, andmedium-to-high morale (among other factors) is a mis-match with leaders having a high preference for makingdecisions (Zammuto and Krakower 1991).Attempting to accomplish the SAIDT tasks with a

pre-2003 NASA organization would also present severalcontingency misfits. First, managers with low preferencefor delegation and high uncertainty avoidance present amisfit with results-based incentives. This mismatch cre-ates confusion and frustration for manager and employeealike (Miller and Toulouse 1986). Second, the group cli-mate would be incongruous with several organizationalcontingencies. The high-trust, low-conflict, high-morale,and high-leadership credibility of the group climate is amisfit with the separation of jobs into rigid specializa-tions and a highly formalized structure (Zammuto andKrakower 1991).The analysis here clearly shows the potential bene-

fits of the transformation underway at NASA. Currentand future missions require the use of technical exper-tise distributed across the field centers. Yet, developingan organization design capable of integrating this exper-tise is challenged by the cultural imperatives and tradi-tions that exist at each center (research versus mannedspace flight versus unmanned robotic space missions,etc.), the persistent center fiefdom mentality and rival-ries, and the inherent slowness with which bureaucracycan be changed in any large organization. The SAIDTdesign attempts to operate more seamlessly in a OneNASA mode: less intercenter competition, more coop-eration, more use of modern information technology tooperate in virtual teams, and more alignment across thecenters and teams to get the institutional resources (staff,facilities, budgets, etc.) in line and executing toward thePresidential Vision for Space Exploration.

6. One-Year Retrospective: What Workedand What Didn’t

6.1. SAIDT PerformanceProject Constellation managers reviewed SAIDT perfor-mance one year later as part of planning for the nextyear’s operations. This review revealed a number of pos-itive outcomes, as well as ongoing challenges, related tothe SAIDT organization design process.On the positive side, the SAIDT organization (shown

in Figure 3) achieved a high degree of virtuality, or geo-graphically distributed operations coordinated via elec-tronic communications. By drawing on staff from seven

NASA field centers, the SAIDT organization effectivelymatched the skills or disciplines required by the taskswith those possessed by NASA staff. The core lead-ership planning team used electronic communicationsto plan the studies, coordinate activities, and achieveteam consensus whenever practical. This structure alsoproved to be flexible and able to take on unexpectedwork effectively.The SAIDT leadership decided to reduce the number

of planned analysis cycles, as recommended by Case 3(reported in §5.4). This certainly helped, as planningthe analysis study tasks to be performed during the firstfull cycle proved to be more challenging, and thus tooklonger than originally thought. The process of pollingeach of the 14 IDTs for their study task proposals, pri-oritizing them in light of programmatic requirements,documenting them into a written and approved studyplan, finding and assigning task leads, and populatingtheir study teams, proved to be challenging even withthe experienced SAIDT leadership and virtual manage-ment tools discussed previously. The high degree of dis-tributed matrix teaming in the overall IDT structure,together with this being the first time through the plan-ning process in the new team structure, contributed tothe longer planning period required.Ongoing challenges remain as well. A key challenge

in any matrix organization is maintaining open horizon-tal and vertical communication. The SAIDT experienceduring the first year of operations underscored this fact.There were instances of miscommunication within theSAIDT planning team that resulted in conflicting guid-ance being issued to lower tiers in the team. The geo-graphically distributed nature of the personnel on theteam was a contributing factor. It should be noted thatthis arrangement of chair and deputy coming from dif-ferent centers was an intentional One NASA feature ofthe original IDT design. The SAIDT leaders resolvedthis issue by communicating more frequently, which alsoillustrates two key factors in the success of implement-ing any organization: individual commitment and action.These are critical to maintaining communication andinformation flow in geographically distributed organiza-tions like the SAIDT.

6.2. Cultural Impact of Organization DesignSimulation Tools at NASA

As mentioned above, the CAIB report that identifiedorganizational and cultural factors as contributing causesto Columbia’s loss provided motivation for the ODT’sresearch into improved organization design and simula-tion methods. Early results from the ODT’s work havebeen promising. First, the effort to apply DSM to designthe IDT structure was honored with a One NASA PeerAward. The award citation read:

In recognition of your achievement in the demonstra-tion of One NASA behaviors. You are being recog-nized by your peers for successfully creating the optimal


Constellation Systems Integrated Discipline Team (IDT)structure by active engagement of all Centers, focusingon the common good of the Agency, and using innovativeorganizational design techniques.

Second, senior NASA managers requested that the ODTapply the techniques outlined in this paper in a follow-on study to simulate the entire Constellation SystemEngineering and Integration organization. Results of thislarger scale simulation would then be used to refine thebroader organization design to improve its performancein future government fiscal years.To build on the success and management recognition

of these initial studies, the ODT continues to look forAgency programs and projects in their formulation phasethat can benefit from organization design and simula-tion. Knowledge capture and broader dissemination tocustomers and prospective users of the tools and meth-ods is a necessary next step. Toward that end, the ODTformalized this activity into an initiative entitled POMS,which captures the knowledge developed over the pasttwo years into a toolkit. The toolkit is a desktop soft-ware application being developed by the ODT to provideprogram and project managers, system engineers, andcost analysts with immediate access to the POMS tools,methods, case studies, and supporting databases throughan intuitive graphic user interface.However, the design of new organizations within

NASA is still being performed by senior managers asa largely intuitive process informed primarily by pastexperience and often influenced by the external operat-ing environment and interpersonal networks. Managersview design of their organizations as a responsibilityand a right that comes with the job. The challenge forthe ODT is to continue the process of institutional-izing the POMS design approach, so that more man-agers are aware of these tools and methods. The goalis not to replace intuitive and experiential-based orga-nization design, but to use it as a starting point forsubsequent, analysis-based assessments of the candidatedesigns, the results of which can either validate the man-ager’s intuition or provide a counterintuitive result thatmerits closer examination.

7. Discussion and ConclusionsThis paper reports on the process of designing a neworganization unit within NASA. The NASA design chal-lenge is to obtain a good or satisficing (Simon 1955)design, where coordination among the distributed unitsin a large organization is paramount to meet the require-ments of time within the available resources and atacceptable quality levels. The new organization mustalso fit within an ongoing, agencywide transformationin operations and organizational culture, while achiev-ing the integration of complex space and ground systemsinto the entire Constellation System. Several themesemerge from this effort.

7.1. The Context MattersOne theme that emerges is that organization designs areshaped by the constraints of the larger organization andthe organizational environment (Pettigrew 1987, Galunicand Eisenhardt 1994). For example, the transformationunderway at NASA seeks to change a bureaucracy, a cul-ture, and traditions that formed over many years. Chang-ing such an organization is also a long-term project. TheSAIDT design team received support from senior man-agers in pushing the boundaries of typical NASA orga-nizational approaches to operate with more cooperationand less competition, using information technology toolsto operate an effective virtual team. This support maynot have been as pronounced had the larger organizationchange effort been absent.In addition, the SAIDT design is constrained by the

technology. Human space exploration requires the devel-opment of a large number of high reliability, interac-tive systems, many of them novel. This work, in turn,requires personnel distributed across many geographi-cally dispersed centers to integrate their work. Thus,simple designs with low coordination would not belikely to accomplish the required tasks.The organization design simulation tools we used fit

both the problem and the NASA culture. Because NASArelies heavily on simulations in the engineering of com-plex technical systems, tools that simulate the humansystems fit the engineering culture that pervades theorganization. Because the process and tools were accept-able to NASA personnel, the result or outcomes weremore likely to be accepted as well, contingent upon cus-tomary verification and validation of the tools’ under-lying algorithms and assumptions. DSM, OrgCon, andSimVision modeling processes all rely heavily on teammember input. This has two advantages: (1) that themodels are informed by the subject matter experts and(2) the models promote “buy-in” and acceptance of theresults. The tools also provide a “technical grammar”(Argyres 1999), which helps the design team generateand evaluate possibilities.This implies that much like the contingent relation-

ships between the environment and aspects of the orga-nization design, there may be a contingent relationshipbetween the tools employed for organization design andthe organization itself. This is not an area that the aca-demic organization design literature has focused on.

7.2. The Tools MatterOrganization design is fundamentally an experience-based process performed intuitively by leaders and man-agers (Hedberg et al. 1976). Managerial intuition islikely to be sufficient in situations where there arefew variables with less interaction than the situationdescribed here. The complexity of the NASA challengeargues for a set of tools and processes that complementmanagerial intuition.


The tools and processes employed here clarified keyissues and questions, challenged assumptions, shapedthe generation of alternatives, and provided a commonframework for consideration of alternatives. For exam-ple, the DSM methodology focuses on the relationshipsbetween project components to facilitate the design ofa modular system architecture or organization structure.OrgCon focuses on the overall fit between organiza-tional components and environmental factors, includingthe organization’s strategy and its environment. Finally,SimVision provides a detailed, information-processing-based view of tasks, processes, and people that allowsmanagers to focus on the effect of “hidden” worksuch as rework, coordination, and wait time on projectperformance.

7.3. Organization Design Process Implications forContingency Theory

The SAIDT design team eliminated misfits as they iter-atively improved the organization design to a satisficingconclusion. The first phase of the project used the DSMto avoid misfits by grouping interdependent project ele-ments together to create 14 integrated discipline teamsthat cut across NASA centers, projects, and functionalareas. The creation of groups with higher degrees ofexternal interactions compared to internal interactionswould have been a misfit (Simon 1962). The secondphase developed the structural properties of the SAIDT,in particular, using OrgCon to assess whether any mis-fits existed between characteristics of the organizationand its environment. Finally, the ODT used SimVisionin the third phase as a means to develop the microlevel,task-oriented SAIDT design. The misfits of the baselinemodel (the likelihood that the project would not meetschedule and budget constraints) generated the search foralternative organizational models. Thus the design teamconcentrated on the alleviation of misfits, rather than thesearch for fit.Yet, organizational contingency theory is more often

presented as a theory of fit, typically between envi-ronmental factors and organization design elements. Acontingency fit is defined as a match “between theorganization structure and contingency factors that has apositive effect on performance” (Donaldson 2001, p. 7),“while a misfit produces a negative effect on organi-zational performance” (p. 15), where the misalignmentbetween organizational and environmental factors leadsto diminished performance (Naman and Slevin 1993,Donaldson 2001). However, misfits are “less well devel-oped in terms of operational statements and empiricaltests” (Burton et al. 2002, p. 1480).Thus, one conclusion from this perspective is that

theory, which is articulated in terms of misfits speaksmore directly to managers than theory, which is statedin fit terms. Managers can use misfits as a mechanism in

the design process. In short, misfits provide the bridgebetween practice and theory.A second implication for theory is that organizational

contingency theory is not completely developed andtested as a misfit theory (Burton et al. 2002); there ismuch to do here. Misfit concepts are complementaryto fit arguments, but that does not mean it is a simpletask of restatement. Given the number of organizationaland environmental elements, there are a vast number ofpotential dyadic interactions, and thus a vast number ofpotential misfits. However misfits are not likely to allbe of the same importance, either in practice (for exam-ple, some were ignored by the design team) or in theory,where some misfits are likely to have greater effect onperformance than others. That is, there is a degree ofperformance loss, which varies across the misfits thatneeds to be developed and understood.

7.4. Limitations and Directions for FutureResearch

This study is limited in its focus. These limitations, how-ever, help point the way for future work. First, this studyexamines the work of a single organizational designteam. Much of the detail work was focused on the devel-opment of the Constellation SAIDT, in particular. Thusthe description here concerns the issues encountered,as well as the tools and approaches used in this onecase. The current pilot program was intended as a proofof concept for NASA, and it is being extended there.Nonetheless, the processes and tools are specific.Second, the single case raises issues of generalizabil-

ity and transferability. NASA is a complex peculiar orga-nization. Few organizations match it. Yet, it containselements familiar to many organizations—the need tomatch strategy, structure, and processes, the challengeof coordinating the work of geographically distributedexperts, and the limitations of organizational context.Thus it provides a worthwhile test bed to examine howfit is achieved in the face of constraints (Siggelkow2001).Finally, we believe that the organization design case

study detailed here may be viewed as an iterative pro-cess of misfit identification and elimination. Our theoriesare often not operationalized or tested with a focus onmisfits. Doing so in the future will more directly addressmanagerial needs and fit the processes they employ.

AcknowledgmentsThis paper’s authors are the SAIDT chairperson, two mem-bers of the SAIDT, and two consulting academics. The authorsthank Darrell Branscome of SAIC, Todd Denkins and JohnKorte of NASA Langley, Kevin Langan of SAIC, DeborahNeubek of NASA Johnson, Borge Obel of Aarhus School ofBusiness, four anonymous reviewers, and the special issue edi-tors for their valuable assistance.


ReferencesArgyres, N. S. 1999. The impact of information technology on coor-

dination: Evidence from the B-2 “stealth” bomber. Organ. Sci.10(2) 162–180.

Baligh, H. H., R. M. Burton, B. Obel. 1996. Organizational consul-tant: Creating a useable theory for organization design. Manage-ment Sci. 42(12) 1648–1662.

Brooks, F. P. 1974. The Mythical Man-Month: Essays on SoftwareEngineering. Addison-Wesley, Reading, MA.

Burton, R. M., B. Obel. 2004. Strategic Organization Diagnosis andDesign: The Dynamics of Fit, 3rd ed. Kluwer Academic Pub-lishers, Norwell, MA.

Burton, R. M., J. Lauridsen, B. Obel. 2002. Return on assets loss fromsituational and contingency misfits. Management Sci. 48(11)1461–1485.

Columbia Accident Investigation Board Report. 2003. Volumes I–VI.http://www.caib.us/news/report/default.html.

Donaldson, L. 2001. The Contingency Theory of Organisations. SagePublications, Thousand Oaks, CA.

Galunic, D. C., K. M. Eisenhardt. 1994. Renewing the strategy-structure-performance paradigm. B. Staw, L. Cummings, eds.Research in Organizational Behavior. JAI Press, Greenwich, CT,215–255.

Greenwood, R., C. R. Hinings. 1988. Organization design types,tracks, and the dynamics of strategic change. Organ. Stud. 9(3)293–316.

Hedberg, B. L. T., P. C. Nystrom, W. H. Starbuck. 1976. Camping onsee-saws: Prescriptions for a self-designing organization. Admin.Sci. Quart. 21 41–65.

Kim, J., R. M. Burton. 2002. The effect of task uncertainty and decen-tralization on project team performance. Computational Math.Organ. Theory 8(4) 365–384.

Levitt, R. E., G. P. Cohen, J. C. Kunz, C. I. Nass, T. Christiansen,Y. Jin. 1994. The “virtual design team”: Simulating how orga-nization structure and information processing tools affect teamperformance. K. M. Carley, M. J. Prietula, eds. ComputationalOrganization Theory. Lawrence Erlbaum Associates, Hillsdale,NJ.

Levitt, R. E., J. Thomsen, T. R. Christiansen, J. Kunz, Y. Jin, C. Nass.1999. Simulating project work processes and organizations:Toward a micro-contingency theory of organizational design.Management Sci. 45(11) 1479–1495.

Miles, R. E., C. C. Snow. 1978. Organizational Strategy, Structure,and Process. McGraw-Hill, New York.

Miller, D., J.-M. Toulouse. 1986. Chief executive personality and cor-porate strategy and structure in small firms. Management Sci.32(11) 1389–1409.

Naman, J. L., D. P. Slevin. 1993. Entrepreneurship and the concept offit: A model and empirical tests. Strategic Management J. 14(2)137–153.

National Aeronautics and Space Administration. 2004. The Vision forSpace Exploration. http://www.nasa.gov/pdf/55583main_vision_space_exploration2.pdf.

Nystrom, P. C., B. Hedberg, W. H. Starbuck. 1976. Interacting pro-cesses as organizational design. R. Killman, L. Pondy, D. Slevin,eds. The Management of Organization Design. North-Holland,New York, 209–230.

Pettigrew, A. 1987. Context and action in the transformation of thefirm. J. Management Stud. 24(6) 649–670.

Pettigrew, A. M. 1992. The character and significance of strategyprocess research. Strategic Management J. 13(Winter) 5–16.

Siggelkow, N. 2001. Change in the presence of fit: The rise, thefall, and the renaissance of Liz Claiborne. Acad. Management J.44(4) 838–857.

Simon, H. A. 1955. A behavioral model of rational choice. Quart. J.Econom. 69 99–118.

Simon, H. A. 1962. The architecture of complexity. Proc. Amer. Phi-los. Soc. 106 467–482.

Starbuck, W. H., F. J. Milliken. 1988. Challenger: Fine-tuningthe odds until something breaks. J. Management Stud. 25319–340.

Steward, D. V. 1981. The design structure system: A method formanaging the design of complex systems. IEEE Trans. Engrg.Management 78(3) 71–74.

Van de Ven, A. H. 1992. Suggestions for studying strategy pro-cess: A research note. Strategic Management J. 13(Summer)169–188.

Weick, K. E. 1977. Organization design: Organizations as self-designing systems. Organ. Dynam. 6 31–46.

Zammuto, R. F., J. Y. Krakower. 1991. Quantitative and qualitativestudies in organizational culture. Res. Organ. Change Develop-ment 5 83–114.

DesigningaNewOrganizationatNASA: … · 2015-07-29 · 2.1....

Documents

Transcript of DesigningaNewOrganizationatNASA: … · 2015-07-29 · 2.1....