PLEASE SCROLL DOWN FOR ARTICLE

This article was downloaded by: [Indiana University Libraries]On: 9 February 2011Access details: Access Details: [subscription number 788830788]Publisher Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Information Systems ManagementPublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t768221794

Continental Airlines Continues to Soar with Business IntelligenceBarbara H. Wixoma; Hugh J. Watsonb; Anne Marie Reynoldsc; Jeffrey A. Hofferd

a McIntire School of Commerce, University of Virginia, Charlottesville, Virginia, USA b Terry Collegeof Business, University of Georgia, Athens, Georgia, USA c Continental Airlines, Houston, Texas, USAd School of Business Administration, University of Dayton, Dayton, Ohio, USA

To cite this Article Wixom, Barbara H. , Watson, Hugh J. , Reynolds, Anne Marie and Hoffer, Jeffrey A.(2008) 'ContinentalAirlines Continues to Soar with Business Intelligence', Information Systems Management, 25: 2, 102 — 112To link to this Article: DOI: 10.1080/10580530801941496URL: http://dx.doi.org/10.1080/10580530801941496

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial orsystematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply ordistribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae and drug dosesshould be independently verified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directlyor indirectly in connection with or arising out of the use of this material.

102

Information Systems Management, 25: 102–112Copyright © Taylor & Francis Group, LLCISSN: 1058-0530 print/1934-8703 onlineDOI: 10.1080/10580530801941496UISM

Continental Airlines Continues to Soar with Business IntelligenceContinental Airlines Continues to Soar with BI Barbara H. Wixom1, Hugh J. Watson2, Anne Marie Reynolds3, and Jeffrey A. Hoffer4

1McIntire School of Commerce, University of Virginia, Charlottesville, Virginia2Terry College of Business, University of Georgia, Athens, Georgia

3Continental Airlines, Houston, Texas4School of Business Administration, University of Dayton, Dayton, Ohio

Abstract As the business intelligence industry matures, it is increasingly important to investigateand understand the nature of mature data warehouses. Although data warehouse research is prevalent,existing research primarily addresses new implementations and initial challenges. This case study ofContinental Airlines describes how business intelligence at Continental has evolved over time. It identifiesContinental’s challenges with its mature data warehouse and provides suggestions for how companiescan work to overcome these kinds of obstacles.

Keywords data warehousing, business intelligence, maturity model, case study

We have a very mature data warehouse at Continental—we havebeen doing business intelligence for a long time, about a decade.New ideas for leveraging the warehouse keep coming up as wecontinue to evolve. Many are not the kinds of major applicationsthat would have justified our initial investment in warehousing.But, these are the kinds of things that create a constant return onour investment.

– Anne Marie Reynolds, Data Warehousing Director,Continental Airlines

Much has been written about the potential for data ware-housing; how to build a data warehouse; and how somecompanies are benefiting from business intelligence; butmost of the writings have been focused on companies’initial experiences. Data warehousing is known to be ajourney, not a destination. Although it clearly is impor-tant to start off strongly, it is equally important to knowhow to move ahead over time. Managers need to under-stand how to evolve data warehouse initiatives to meetthe changing and growing needs of the business. Compa-nies need to continuously extract value from their signif-icant on-going warehousing investments.

On a journey, if you know what lies ahead, the more aptyou are to overcome obstacles and take the most satisfyingroute. In data warehousing, many folks are blind to whatlies ahead when they begin. We have much to learn

about the opportunities and challenges posed by a maturedata warehouse. How do early decisions, such as datamodeling technique or platform selection, impact latercapabilities? What mechanisms best enable a warehouseto expand its reach and business impact? What obstaclesdo warehousing teams need to overcome over time?Answers to these questions may better guide even themost adventuresome of warehousing journeymen.

To explore these questions, the authors conducted aseries of interviews with business intelligence users andtechnologists at Continental Airlines. The airline has hada data warehouse in place since 1998 and is considered aleader in business intelligence (Watson, Wixom, Hoffer,Anderson-Lehman, & Reynolds, 2006). Over time, thewarehouse has grown and evolved in exciting ways,spreading out across the enterprise’s business areas andgeographical locales. Applications have moved frombeing strategic and tactical in nature to ones that arehighly operational. The numbers of users, applications,and data increase annually.

In this article, we begin by discussing the characteris-tics of a mature data warehouse. The characteristics aredrawn from the maturity or stage models that have beendeveloped for data warehousing. Next, we describe thecurrent state of Continental Airlines’ data warehouse,highlighting evidence of the warehouse’s maturity.Then, we discuss three ways in which Continental hasexpanded the reach of its data warehouse, and weexplore the mechanisms that enabled the expansion. We

Address correspondence to Hugh J. Watson, Terry College ofBusiness, University of Georgia, Athens, Georgia 30602, USA.E-mail: hwatson@uga.edu

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

Continental Airlines Continues to Soar with BI 103

conclude with challenges that Continental now faceswith its mature data warehouse and with some sugges-tions for how companies can work to overcome thesekinds of obstacles.

Mature Data Warehouses

An organization does not develop a mature data ware-house overnight; rather it is the result of a progressionthrough a series of earlier stages. It is rare that a ware-house skips stages. Each stage builds on a previous stageand provides a greater set of capabilities. A “mature”warehouse is one that has evolved to the point that itis part of the institutional fabric and integral to thefunctioning of the organization.

In order to better understand mature data warehous-ing, it is useful to step back and look at antecedents ofthe concept. Maturity, stage, and evolution models, asthey are variously called, have been popular in disci-plines for many years. The fundamental concept is thatthings change over time, in sequential, predictable ways.Stage models have been used to describe, explain, andpredict organizational life cycles, product life cycles, andbiological growth. They also have been popular in theinformation systems field. The classic model is Nolan’sStages of Growth Model that identifies specific character-istics that firms demonstrate as their IT departmentsmature (Nolan, 1979).

Though various stage models may differ in terms ofthe number of stages and what the stages are called, theyare all similar in that they break down a phenomenon’sevolution into a series of distinct phases with characteris-tics associated with each phase. For example, Nolan origi-nally had a four stage model (i.e., Initiation, Expansion,

Formalization, and Maturity) that he later expanded to asix stage model to reflect advances in practice.

Each stage model is composed of component stagemodels, which collectively identify the stage of thehigher-level model. For example, Nolan’s model includesdimensions of application growth, personnel specializa-tion, and management techniques. Possibly, each ofthese components reflects a different growth stage; appli-cation growth may be mature, while personnel special-ization is expanding. However, overall and collectively,the components define a stage of evolution.

The data-warehousing field also has maturity models.Watson, Ariyachandra, and Matyska (2001) suggest thatbusiness intelligence has Initiation, Growth, andMaturity stages. Eckerson (2006) proposes a five-stagemodel (i.e., Prenatal-Infant, Child, Teenager, Adult, andSage). Each of these models uses characteristics or dimen-sions in categorizing the stages. For example, Watson,et al categorize the stages based on a data warehouse’sdata, stability of the production environment, ware-house staff, users of the warehouse, impact on users’skills and jobs, use of the warehouse, organizationalimpacts, and costs and benefits. Eckerson uses scope,funding, warehouse staff, governance, standards, archi-tecture, executive perception, data latency, and businessintelligence focus. Together, the models provide a com-prehensive list of characteristics to use in categorizingvarious maturity stages; see Figure 1 for a list of charac-teristics of data warehouses in mature stages.

The research around these models teaches us thatmaturity is an elusive state. Often, maturity means thatan organization repeats and likely improves uponwhat has worked well in the past in new areas of theorganization. Maturity means that the initiative is partof the culture. There is learning from success and failure,

Figure 1. Characteristics of mature data warehouses.

Eckerson Maturity Model Adult Stage

Watson et al Maturity Model Maturity Stage

• Scope. Enterprise. • Data. Enterprise, well integrated, for multiple time periods.• Architecture. Warehouse, with dependent data marts.• Stability of the Production Environment. Procedures are routine

and documented.• Warehouse Staff. Experienced in-house personnel, with well-defined

roles and responsibilities.• Users. Organization-wide, suppliers and customers may be users.• Impact on Users’ Skills and Jobs. Users throughout the organization

need improved computer skills to perform their jobs.• Applications. Reports, predefined queries, ad hoc queries, DSS, EIS,

data mining, and integration with operational systems.• Costs and Benefits. Benefits include timesaving, new, and better

information, improved decision-making, redesigned business processes, and support for corporate objectives.

• Organizational Impact. Organization-wide and often strategic, tactical, and operational.

• Funding. IT/Business.• Team. Corporate IT.• Governance. Business Intelligence Stewardship Team.• Flexibility/Standards. Plan global, act local.• Architecture. Enterprise data warehouse.• Type of System. Strategic System.• Executive Perception. Drive the business.• Data Latency and Freshness. Low latency, high freshness.• Business Intelligence Focus. What should we do?• Business Intelligence Output. Action.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

104 Wixom et al.

but also there is a perception of enough success that theinitiative is allowed to continue and grow, until someradical event or new opportunity makes the initiativeobsolete.

According to the characteristics listed in Figure 1,Continental Airlines has a mature data warehouse. Therest of the article will describe the warehouse, with afocus on its maturity in terms of the growth of its appli-cations and usage. As data warehouses mature from anapplication perspective, their scope and reach increase.The warehouse delivers the most direct business value bymaturing in this way.

Continental’s Mature Data Warehouse

Continental’s original data warehouse team was formedin 1998 and charged with integrating a diverse set of datafrom the key mainframe source systems throughout thecompany onto one platform. Two specific areas drove theoriginal project: revenue management and marketing.

The Revenue Management department is responsiblefor determining the number of airlines seats offered at aparticular price point (i.e. booking class) for every Conti-nental flight. In 1998, they needed to capture detailedbooking data at the lowest level of detail for analysis.Improvements in the ability to understand passengerbehavior in order to optimize pricing decisions wereexpected to result in significant revenue opportunitiesfor the airline.

At about that same time, the Marketing departmentneeded to understand customers better. The existing fre-quent flyer system measured customer activity based onmiles flown, but the Marketing department wantedinsight into the revenue associated with each customer.They also wanted to know which customers wereaffected by flight delays and cancellations and other ser-vice interruptions. Understanding customers better wasexpected to improve Continental’s profitability and itsability to respond to customers’ needs, thus increasingcustomer loyalty.

A Teradata platform was chosen, primarily for scal-ability and ease of administration reasons. Even thoughtwo specific departments were funding the initial invest-ment, the executive sponsors had the vision to conceiveof a data repository that would ultimately encompass theentire population of corporate data required to supportdecision making, so scalability was critical.

Continental recognized the need and planned for real-time business intelligence at the outset of its data ware-housing initiative. The warehouse group built a dataacquisition infrastructure whereby source systems wouldfeed a queue in either batch mode or in a constant flow.A warehouse loading process continuously monitored datasources and pulled data into the warehouse. The initial

feeds of data were all performed in batch. Currently, fif-teen percent of the subject areas are loaded continu-ously, yet the original data acquisition infrastructure isstill a highly effective loading mechanism.

The data within the warehouse was modeled using astrict third normal form modeling technique, and thedata used standard naming conventions. A dedicatedand strong-willed database administrator ensured thatthere were few exceptions to these policies over time,regardless of short-term pressures. The intent was to cre-ate an underlying data layer that was not shaped byunique needs, but instead was kept “vanilla” so that itcould easily evolve over time to meet a wide breadth ofrequirements.

The general philosophy was to grant users access to alldata, unless there was a reason not to do so. Thus, whenusers received a warehouse ID, they immediately wereable to access a common enterprise-wide data layer. Somedata in the data warehouse, such as credit card numbersand employee salaries were restricted, for obvious reasons.This “open data” philosophy remains intact today.

By 2002, the Revenue Management and Marketing ini-tiatives were successful, more than offsetting the datawarehouse investment (see Anderson-Lehman, Watson,Wixom, and Hoffer, 2004). Both functional areas workedwith the data warehouse team to calculate and docu-ment the most tangible and significant benefits, whichestablished credibility for future funding.

Over the nine years that the data warehouse has beenin existence, the scope has grown to include more than50 subject areas and 1400 named users writing ad-hocqueries. In addition, the data warehouse is the “singlesource of the truth” for 70 applications developed byContinental. These applications fall into three generalcategories (see Figure 2).

In 2007, the Continental data warehouse had trulyachieved a global scope. About half of the user commu-nity accesses the warehouse from the company head-quarters in Houston, while the other half is spread across75 cities throughout the world. The data warehouse staffhas conducted international training sessions in London,Guam, and Tokyo, which are some of the airline’s centersfor international sales and pricing; and there are plans tovisit Latin America. These user groups are becoming datawarehouse analysts, rather than simply consumers ofprewritten reports.

A team of 15 people supports the Continental datawarehouse, and this group is responsible for thedata transformation development, application interfacedevelopment, user support and training, database admin-istration and production support. The 24 × 7 productionsupport is shared by all of the members of the group androtated weekly. Production incidents are reviewed eachweek to ensure that the root cause of each incident isaddressed.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

Continental Airlines Continues to Soar with BI 105

The data warehousing team size remains relativelysmall because application development is primarily donewithin the functional areas. Each department has at leastone fairly technical employee who serves as a liaisonbetween the warehouse group and the department. Thisemployee ensures that the appropriate business needsare communicated to the warehouse group; he or shebuilds warehouse queries, reports, and/or applicationsfor use within the department.

Initially, a steering committee prioritized warehouse ini-tiatives and actively managed the warehouse evolution;however, once the enterprise data foundation was firmly inplace and new projects became more focused on niche func-tional needs, the steering committee evolved into an advi-sory body. The warehouse management now prioritizesprojects, with input from steering committee members.Annually, the warehouse management presents the coredevelopment projects to upper management for approval.

The funding model for the warehouse also changedover time. Initially, areas like revenue managementand marketing built a business case for the warehouseand provided its funding. Each time the warehouserequired a significant upgrade to the infrastructure, adifferent functional area would make the necessaryinvestment although all areas would benefit. A signifi-cant shift occurred recently when the warehouse direc-tor secured a multi-million dollar capital investment forthe warehouse through the IT budget. The belief is thatthe warehouse has become a critical part of the IT infra-structure and should be funded as such.

Application Maturity

Over time, Continental’s data warehouse applicationportfolio has grown, and with this growth Continentalhas realized considerable business value. The airlinehas generated warehouse application growth in three

unique ways: by constantly adding new business groupsto the warehouse user base, with new applications tomeet their needs; widening the user base across theglobe; and integrating warehouse capabilities into opera-tional business processes. The following sections provideexamples of how each type of expansion has occurred.

Enterprise Expansion

One way to expand data warehouse applications andusage is by increasing the number of departments orsub-groups that use the warehouse. Initially, the revenuemanagement and marketing groups championed thedevelopment and use of the warehouse. Over time, usagespread to other groups, such as Human Relations andTechnical Operations. The following sections describehow this expansion occurred.

Payroll Department

Continental, like many organizations, has legacy systemsthat are cumbersome to use for querying and reporting.Processes that require data from legacy applications oftenare time-consuming and require manual intervention.For example, prior to the data warehouse, a person in thePayroll Department was responsible for ensuring thatthe amount Continental is billed for benefits matches theamount of benefits actually being used by its employees.Each Monday, this employee received a benefits reportand then worked all week to reconcile the numbers.

Being a part of the overall Human Relations (HR) Depart-ment, Payroll had watched the HR area leverage the datawarehouse, and over time, the Payroll Department beganto realize they, too, could tap into warehouse data for theirown needs. At first, paycheck information was loaded intothe data warehouse. The group created simple queries

Figure 2. Types of applications at Continental.

The application is a recipient of data provided through a regularly scheduled batch process.

For example, the warehouse provides revenue data to a Flight Profitability System and a Travel Agency management system, called Sales Insight.

The application is the recipient of data provided in real-time through a web service or direct query to the data warehouse database.

For example, the warehouse provides a real-time look at today’s flight operations through a web interface that includes the capability to drill-down to flight detail.

The application resides on the data warehouse. The warehouse performs a predefined function using the data, usu-ally from several data sources.

For example, the data warehouse performs statement processing for the small business Reward One program, combining frequent flyer and revenue information with the Reward One profiles.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

106 Wixom et al.

against that data to review payroll data. This was easy forPayroll as the users were already familiar with the data.“Once we were comfortable with that, we started realizingwe could get a lot more out of the data warehouse. That iswhen we started thinking of having the data warehouse dosome payroll calculations for us.” Currently, the ware-house group is designing a data warehouse report to per-form benefit deduction reconciliations against YTDcoverage information from Continental’s Benefits Provider.This one task will free up an employee to spend his weekdoing more value-added tasks for the department.

And, there are additional plans for the data warehouse.Payroll is working to transfer time and attendance data tothe warehouse to meet more unique departmental needs.The current goal is to be able to input an employee num-ber and immediately have access to all related HR data sothat the group can answer employee questions within theduration of a phone call. Once time and attendance dataexist in the warehouse, Mullane predicts that the com-pany as a whole will have the capability to even betterunderstand what happened at the airport on a specificdate. “The more data we put in for ourselves, it seems likeother divisions get more and know more about howthings are happening at the company level.”

Reliability Engineering Department

Reliability engineers currently are moving their pro-cesses from a legacy COBOL Focus system to queries andapplications based on the data warehouse. Reliabilityengineers ensure that the Continental fleet operatesreliably, and they work to reduce delays, cancellations,and aircraft returns. Although their previous Focus sys-tem was highly useful, it was extremely costly to use.EDS maintained the system as an outsourcer, so eachtime the engineers asked for a report or query in thepast, they had to write a check to EDS. Continental as acompany began to push the group to lower costs; thedata warehouse offered the engineers a much bettervalue proposition, with expected savings of about$100,000 annually.

The engineers also anticipate that the warehouse willincrease their capabilities. The Focus tables had a very nar-row scope; the warehouse allows reliability engineers tojoin tables like never before possible. The engineers sus-pect that having access to a larger scope of enterprise datamay allow them to tap into special needs, some that cur-rently may be unknown. “The warehouse should allow usto think outside the box . . . or think into a different box.”

Although the reliability engineers are still in transi-tion over to the warehouse, they have used it enough tobelieve in its value. Recently, Reliability Engineeringreceived a call from a related area within Technical

Operations, inquiring about options for storing theirdata. “We recommended that they go with the data ware-house,” explains a Sr. Systems Analyst. “The warehousegroup is knowledgeable and easy to communicate with.They generally recognize what our role is at Continentaland what our needs for the data are. They provide dataaccess pretty rapidly.” The data warehouse group’s highlevel of service and close working relationship with Reli-ability Engineering, along with the warehouse’s valueproposition, will encourage the reliability engineers torecommend that others also leverage the data warehouse.

Enterprise Expansion

Continental continuously adds new organizationalgroups and users to the warehouse, and this happens ina fairly predictable way (see Figure 3). First, an organiza-tional group observes a related area using the datawarehouse to do things easier, cheaper, or better. Typi-cally, the groups rely on similar kinds of data and sys-tems; therefore, it is reasonable to deduce that if therelated area is leveraging the warehouse, then the newgroup can derive value from the warehouse as well.Next, the group begins by asking simple data queries ofknown data. This important step builds user skills, con-fidence, and trust in the data warehouse. Finally, thegroup fully leverages the warehouse by developing cus-tom applications based on the warehouse and by desig-nating a liaison within the group who will ensure thatbusiness needs align with the data and capabilities ofthe warehouse.

Global Expansion

A second way to expand data warehouse applicationsand usage is by rolling out applications globally. Theinitial warehouse applications were headquarters-cen-tric; they best met the needs of employees based inHouston. At times, these same applications also servedglobal needs; but this was not always the case. The

Figure 3. Enterprise expansion.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

Continental Airlines Continues to Soar with BI 107

following sections explain how Continental was able toengage international areas to leverage warehousecapabilities.

Revenue Management, Japan

Aaron Sacharski, Continental’s manager of pricing andrevenue for Asia Pacific, began working for the airline inits Headquarters Revenue Management group in 2002.When Aaron arrived, Revenue Management was heavilyinvolved with the data warehouse, using it to make microand macro types of decisions. The warehouse enabled impor-tant capabilities; its value to the Revenue Managementgroup was obvious. During this time, he met people inthe data warehouse group and built up his own queryingand data analysis abilities.

In 2005, Aaron moved to Japan to manage pricingand revenue for Asia Pacific, fully intending to con-tinue leveraging the data warehouse in his new role.Two obstacles hindered his progress. First, the techni-cal environment in Asia was not as mature; bandwidthwas narrow, and software and hardware wereexpensive.

The second and most challenging hurdle was thatthe data, queries, and analytics that worked at head-quarters were not applicable to the Asian market. Forexample, in Continental’s domestic market, there is asmooth booking curve that can be analyzed to predictflight demand behavior. In Japan, there is a high per-centage of travel agent bookings, and agents are notrequired to migrate their bookings to the Continentalreservation system until 30 day prior to departure.Thus, until a month in advance, analysts have minimalknowledge of flight bookings; it is nearly impossible tocreate a demand curve based on current flight bookingdata.

Another reason that Japan analysts could not effec-tively use data, queries, and analytics from headquarterswas the time difference. Some of the data warehouse rev-enue management tables used to predict future bookingswere populated by batch loads each night at midnight.This was fine for analysts located in Houston, but thedata was not as useful to analysts in Asia who receivedthe data a day and a half later.

For these reasons, the analysts in Japan perceived thatthe warehouse could not meet their unique needs, andthere was little interest to learn about the tool. Aaron,however, had deployable knowledge of the warehousedata acquired in Houston. Although the existing tech-niques from Headquarters did not fit Asia as well, Aaronused his data warehousing knowledge in working withJapan to identify data that could improve pricing andrevenue decisions.

Since 2005, the Japan pricing and revenue managementteam worked in earnest to understand what warehousetables were relevant to Japan, and a warehouse team mem-ber built specialized tables to directly meet their needs.They began to develop queries using PNR (passenger namerecords) data because it is loaded into the warehouse in real-time. At the same time, the technology group for Asiainvested in extra bandwidth for the analysts to improvequerying efficiency. As the warehouse increasingly helpedthe group to make better decisions, usage of the warehousespread.

Tax Department, London

In the United Kingdom, Continental must pay a depar-ture tax for passengers who leave the U.K. on Continentalflights. Each month, employees in the London officecalculated this departure tax by manually reviewing therecords for every passenger who traveled out of London,and the employees submitted the appropriate amount tothe government. If passengers are passing through theU.K. in less than 24 hours, they are exempt from the tax,but the manual process could not always identify thoseindividuals. Thus, Continental regularly overpaid thedeparture tax, which equated to a $300,000 annual costfor the airline.

Last year, several members of the Continental Londonoffice were visiting Houston for routine training, whichincluded a presentation by the warehouse group. Duringthe presentation, the London employees noticed thatdata in the warehouse potentially could identify passen-gers who were exempt from the departure tax. Theyapproached the warehouse team to build a specializedapplication. Now, the group runs a monthly query to thewarehouse, prints out a report with an accurate depar-ture tax amount, and submits the report. The applicationeliminates significant time and overpayment.

Global Expansion

The global spread of warehouse applications begins withthe development of core capabilities in a headquarters orcentral locale. As skills in the base location mature,employees transfer knowledge by physically transferringto global locations or by conducting training programsfor global groups. Seasoned warehouse users can usetheir strong skills to identify gaps between existing ware-house capabilities and unique global needs. Over time,applications can be developed to meet those uniqueneeds. This builds confidence in the warehouse by globalusers and encourages use. Figure 4 illustrates the globalexpansion process.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

108 Wixom et al.

Process Expansion

A third opportunity for expansion includes changingprocesses, looking for ways in which warehouse data canchange the way business is done, making things easierand faster—or doing things differently. Potentially this kindof change can create a competitive edge for Continental.

Reservation Complaint Handling

In the past, when a customer called a Continental reser-vation agent with a complaint, the agent would simplylisten to whatever the customer told them and send theinformation through an antiquated communication pro-cess to the Customer Care department. Members of theeighty-person Customer Care team would then take theinformation, print it, scan it, re-key it into a customercare system, and then work to resolve the case. The pro-cess was time consuming, and it left open the possibilitythat customers could game the system and receive dupli-cate compensation for the same incident. Continental esti-mates that this duplication cost the company $1 millionannually.

Other airlines have addressed this issue by only allow-ing customer complaints via web or email; complaintsare no longer accepted by phone. Continental has a dif-ferent approach:

If the customer is calling us with a complaint, then they want usto listen to them. If we can put in place a process that reduces theoverhead and so streamlines the process that literally there reallyis no reason for us to discontinue telephones as a channel forcustomers to communicate to us, we are better off.

- John Brinker, Senior Manager CRM Strategy andArchitecture

Continental created an automated process that takes avariety of data inputs from the warehouse, runs thoseinputs through a proprietary rules engine, and gener-ates a recommendation on how a particular customer

should be handled (see Figure 5). A reservation agentcan trigger this process while a customer is on thephone, and act on the automated recommendation,which is available within seconds. The automatedprocess eliminates several steps from before, and it hasreduced the customer-care group count by ten people.This new process reduces costs so that Continental cancontinue serving customers in a manner that meetscustomer needs. The customer complaint applicationis now in pilot with 50 reservation agents, and willbe rolled out to the entire Reservations department inQ3 2007.

Flight Performance

Prior to the data warehouse, Continental Operations builtand managed their own information and reporting sys-tems. The systems support staff was very small; when asupport employee went on vacation or was sick, systemswere put on hold until the person returned to work. Even-tually, management mandated a move to the warehouseto improve continuity of the support operations.

Steve Hayes, a manager within this operations sup-port group, has leveraged the warehouse for his area insignificant ways. For example, he has built a real-timestatus application that communicates up-to-the-minuteperformance statistics on how the airline is operating.And, when Jet Blue and American Airlines were criti-cized for incidents that involved stranding passengersin planes for long periods of time (Cummings, 2006;Zeller, 2007), Hayes was able to adapt his application, andhelp Continental avoid similar situations. Continental’sold process for detecting these kinds of events was manualand time consuming. Hayes explains, “You had to huntand peck through flight logs. In the middle of a snowstorm, you don’t have time to do that.”

Once Operations identified the need to monitorplanes on the tarmac, Steve added an alert to the real-time performance statistics application. Now, flights thatsit on the ground away from a gate for at least two hoursimmediately appear on the screen. In real-time, Opera-tions can work to get those flights off the ground, or getthem back to the gate in a timely manner.

The warehouse also has helped streamline Opera-tions reporting processes. In the past, Continentalmanually tracked the reasons for flight delays (e.g.,weather, part failure); there are about a hundred delaycodes. Sometimes stations forgot to record the reasonfor delay, so Operations regularly ran a query on thelegacy system, downloaded the results into Excel,emailed the results to the general managers, whowould then fill in the blanks and send information backby email or telex. According to Hayes, “It would take

Figure 4. Global expansion.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

Continental Airlines Continues to Soar with BI 109

forever to track down the information and update thecodes into the legacy system.”

Using the warehouse, Hayes built an application oneweekend that automatically lists flights that need delaycodes for each station. The general manager now directlylogs into the application, clicks on a flight, and entersthe delay code. The new process eliminates multiplesteps, and creates much more accurate results. Hayesexplains that this situation is representative of how henow can quickly develop simple applications or applica-tion enhancements using the data warehouse that havehigh impact to Operations processes.

Process Expansion

Process change can be difficult because people often arereluctant to change. However, users are much morewilling to embrace change when warehouse applicationsclearly reduce steps, eliminate costs, or save time. Torecognize change opportunities, users must first under-stand the breadth of data that exists within thewarehouse. Then, when key business needs arise, userscan better match these needs to warehouse capabilities;

at that point, they can leverage data for business processimprovement, or even re-engineering (see Figure 6 forthe steps of process expansion).

Maturity Challenges

Maturity does not equate to easy. Thus, once organiza-tions establish mature data warehouses, they should notexpect a guaranteed smooth journey. Although earlyobstacles have been overcome, new challenges emerge as

Figure 5. The warehouse-enabled customer complaint process.

Customer callsContinental

Reservations

Agent requestsresolution

Actual eventdetails

CustomerHistory

CustomerValue

Business Rules Engine Fair Compensation

Agent resolvesrequest

Data Warehouse Customer Complaint Application

Complaint

Complaint ProposedResolution

TicketInformation

Data Warehouse

ActualResolution

Figure 6. Process expansion.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

110 Wixom et al.

a warehouse matures. Below we discuss some of thematurity challenges faced by Continental.

Staffing

Only recently, since the mid-1990s, has database market-ing gained a stronghold in the marketing industry. Thesupply of people who have database marketing skills,such as SQL and query writing, data warehousing, and anappreciation of data has not yet caught up with demand.Continental’s marketing group proactively hires peoplewith those skills; in fact, new marketing employees musttake an SQL test as standard operating procedures. But,the right hires are hard to find. And, once employees arehired and trained, they can be difficult to retain. Onceemployees develop strong analytical and data skills, theycan leave for higher salaries and promotion opportunities.

The staffing problem is not limited to functional areaslike Marketing; the IT group grapples with these sameissues. Talented data warehouse professionals are valu-able, and the turnover in the group has been fairly high.The warehouse management strives to address this byproviding opportunities for cross training and advance-ment where possible, and to keep employees challenged.

Scalability and Performance

Continental currently views technology as a majorhurdle for continued success. For one, applications, suchas the customer complaint application, must performquickly. Reservation agents are uncomfortable if theirsystem does not respond in at least five seconds; thus,much effort has focused on ensuring the data warehousecan deliver fast performance. Global locations increas-ingly need to adapt their technical environments toexpanding usage and more data moving across thenetwork. In short, Continental understands that the datawarehouse must be viewed as an operational tool.The warehouse team invests in significant hardwareupgrades so that the processing power of the warehousesufficiently meets usage requirements.

Business Continuity

Recently, the warehouse group scheduled a systemupgrade that required a window for system outage. Badweather was predicted for that same time period. Opera-tions contacted the warehouse group and asked them toreschedule the upgrade; the warehouse had to be avail-able for use during the bad weather to help Operationsmanage airline performance. The data warehouse wascritical for managing the severe weather event.

The warehouse management acknowledges thatbusiness continuity and disaster recovery are clearlybecoming mandatory now that the warehouse enablesimportant operational processes. It costs a lot ofmoney to maintain an alternate data warehouselocation for an emergency, but if Operations needsthe warehouse to manage a critical weather event,this implies that investments must be made in ware-house disaster recovery. The data warehouse manage-ment is still assessing the alternatives to address theseissues.

Data Volume

The expansion of warehouse usage and applicationsleads to a growing volume of data within the warehouse.Brinker senses that the incredible amount of marketingdata can impede Marketing’s ability to perform true anal-ysis. Large volumes of data are “hard to digest.”

This observation is consistent with the increased inter-est in predictive analytics within the data warehousingcommunity. At the 2006 The Data Warehousing Instituteconference, business leaders ranked predictive analytics astheir number one area for opportunity in business intelli-gence. Predictive analytics may be one solution when datavolumes grow too large for the human mind to manage.

Real-Time Data

Common wisdom says that as users become accustomedto business intelligence, they increasingly want to seedata in real-time. Rarely do we think of global expansionas a driver of real-time data, but that certainly was thecase at Continental. Batch loaded data was typically finefor analysts working in headquarters in Houston, butthat same data did not meet the needs of analysts locatedaround the globe. As organizations roll out applications,they must be aware of the latency needs of their analystsand impact of time zones on the freshness of the ware-house data. It may be that latency plans have to be“refreshed” with global expansion.

Service Delivery

Initially, the warehouse team developed the applicationsfor the warehouse users. The users were unfamiliar withdata warehousing, and they had a hard time conceptualiz-ing the kinds of applications that would be possible. Thus,the warehouse team used a prototyping approach todevelop user applications for them. Sometimes theapproach resulted in applications that were highlysuccessful. Other times, the approach was unable to capture

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

Continental Airlines Continues to Soar with BI 111

the underlying user requirements; users found the applica-tion too hard to use, or not useful for their needs.

As the user skills around the data warehouse matured,the warehouse team moved away from application devel-opment, encouraging functional areas to develop theirown warehouse applications when possible. Currently, thewarehouse team plays much more of an advisory role,traveling around the world to educate users, listen toneeds, and ensure that the warehouse remains relevant tothe user base. This approach is highly effective. Functionalareas perceive that the warehouse team delivers high qual-ity service and responds quickly to their needs.

Conclusions

Although Continental continues to face challenges withits data warehouse, the organization has realized anenviable level of warehouse maturity. Continentalrealizes significant cumulative benefits from the use ofapplications that cross the enterprise, span the globe,and enable fundamental business processes.

Stage models are sometime criticized for failing toexplain how organizations move from one stage toanother. The experiences at Continental, however, pro-vide insights into what facilitates the movement into amature stage, particularly from an applications andusage perspective. Several facilitators make the wide vari-ety, growth, and spread of applications possible. Weconclude by discussing and illustrating these facilitators,each of which are at their own stage of evolution.

A Common Data Foundation

Continental warehouse users advise not to underesti-mate the value of standard naming conventions, a nor-malized data structure, and user-friendly metadata. SteveHayes from Operations explains, “Uniform naming con-ventions across tables is a big deal. This means that I canaccess a table that I have never worked on before, butbecause we have standard naming conventions and aweb-based data dictionary, I can determine the field’spurpose, value, and be able to include it in my queriesand reports.” Warehouse management observes that thedeeper and broader a users understanding of the data,the more creative and usefully the data is applied.

“Open Data” Philosophy

Most companies initially give users limited access to datain the warehouse, likely restricting access to a subjectarea or two most appropriate for their function. At Conti-nental, the philosophy is exactly the opposite. When

users receive a warehouse ID, they immediately haveaccess to all data in the warehouse that does not requirespecial permission. Although only some users can accesshuman resources data or passenger credit card numbers,everything else is fair game.

This philosophy helps generate new uses for the ware-house. According to Brinker, “If I had not had access to thebroad scope of information that was in the warehouse, Iprobably would never have drawn the conclusions that Idid regarding how we could automate the customer com-plaint process.” Brinker explains that in other organiza-tions, a typical marketing user would have visibility onlyinto the company’s customer loyalty program, and he orshe would not interact with operational performance data,check-in information, or passenger name records. BecauseBrinker did have access to the latter information, he wasable to match the known business need of improving cus-tomer care with warehouse capabilities.

A Culture of Data

Continental is fortunate to have a data-driven culturethat was initially developed by former CEO GordonBethune and his management team back in the late1990’s. Employees across the enterprise view data as acorporate asset that can be turned into competitiveadvantage when used properly. Functional areas inten-tionally hire employees who will perpetuate this view.Thus, using the warehouse as a decision support aid isconsistent with the fundamental employee mindset.

Business-IT Hybrids

In some companies, the warehousing staff has strongtechnical skills but limited business knowledge, and thebusiness side has limited technical skills but good busi-ness knowledge. At the intersection of the warehousingand business organizations, there is a dramatic changein the technical/business skills and knowledge mix. AtContinental, liaisons work in each functional areas andmanage the communication between the warehousegroup and business unit. These liaisons are highly techni-cal people with deep functional area knowledge andexpertise. They eliminate disconnects between the tech-nology and business groups and ensure that the rightbusiness needs are met, in the right way.

Final Thoughts

As organizations begin or continue their data warehous-ing journeys, they are advised to assess the current matu-rity of their efforts, and then look ahead to better prepare

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011

112 Wixom et al.

for what is to come. There is a chance that making theright moves now will avoid future missteps. Clearly, Conti-nental provides excellent insights for warehouse journey-men, as the airline soars ahead with business intelligence.

Author Bios

Barbara H. Wixom is an Associate Professor at theMcIntire School of Commerce at the University ofVirginia. (bwixom@mindspring.com)

Anne Marie Reynolds is the data warehouse director atContinental Airlines. (areyno@coair.com)

Hugh J. Watson holds a C. Herman and Mary Virginia TerryChair of Business Administration in the Terry College ofBusiness at the University of Georgia. (hwatson@uga.edu)

Jeffrey A. Hoffer is the Sherman-Standard Register Professorof Data Management at the University of Dayton.(jeff.hoffer@notes.udayton.edu)

References

Anderson-Lehman, R., Watson, H. J., Wixom, B. H., and Hof-fer, J. (2004). Continental Airlines Flies High with Real-Time Business Intelligence. MIS Quarterly Executive, 3(4),163–176.

Cummings, C. (2006, December 30). Passengers Stuck onPlane Over 8 Hours. The Dallas Morning News.

Nolan, R. (1979). Managing the Crises in Data Processing.Harvard Business Review, 57(2), 115–126.

Eckerson, W. (2006). Performance Dashboards: Measuring, Monitoring,and Managing Your Business. John Wiley & Sons: Hoboken, NJ:

Watson, H. J., Ariyachandra, T., & Matyska, R. J. (2001). DataWarehousing Stages of Growth. Information Systems Manage-ment, 18(3), 42–50.

Watson, H. J., Wixom, B. H., Hoffer, J., Anderson-Lehman, J., & Rey-nolds, A. -M. (2006). Real-time Business Intelligence: Best Prac-tices at Continental Airlines. Information Systems Management,23(1), 7–19.

Zeller, T. (2007, February 16). Held Hostage on the Tarmac: Time fora Passenger Bill of Rights? The New York Times.

Downloaded By: [Indiana University Libraries] At: 22:49 9 February 2011