4 2 I E E E S O F T W A R E P u b l i s h e d b y t h e I E E E C o m p u t e r S o c i e t y 0 7 4 0 - 7 4 5 9 / 0 3 / $ 1 7 . 0 0 © 2 0 0 3 I E E E

a firm basis on which to sustain the processimprovement efforts.

At Tata Consultancy Services (TCS), weblended Six Sigma concepts and variousSoftware Capability Maturity Model keyprocess areas1 into our own quality man-agement system. We structured our QMSaround SW-CMM Level 2 and Level 3 re-quirements—including quantitative processand product measurements and analysis,process improvements for defect preven-tion, and process optimization—and rein-forced these with Six Sigma concepts, suchas initiating Six Sigma projects for continu-ous improvements. In this way, TCS-QMSprovides an organized structure for qualityimprovement and customer focus, pavingthe path to CMM Level 5.

Here we describe our approach and itsbenefits, explaining how we leverage SixSigma to gain statistical insight into the align-ment of software goals and customer needs.

We also discuss how our method helps initi-ate the cultural change required to ensurecontinuous improvement and effective imple-mentation of SW-CMM’s Level 4 and 5 KPAs.

TCS’s CMM-based processframework

TCS is India’s largest IT enterprise, pro-viding IT and business consulting services toorganizations in government, business, andindustry in India and abroad. Our servicesare varied and straddle many different indus-tries, including finance and banking, insur-ance, telecommunications, transportation,retail, manufacturing, pharmaceuticals, andutilities. Over time, we’ve integrated variousquality concepts into the TCS-QMS to meetchanging business goals.2

� 1993: Documented software develop-ment practices, maintenance, and con-version to conform to ISO 9000

focusBlending CMM and Six Sigma to Meet Business Goals

Mala Murugappan and Gargi Keeni, Tata Consultancy Services

Integrating SixSigma with SW-CMMprovides theframework forprocessimprovement andcustomer focus byaligning processimprovements goalswith customerexpectations.

Capability maturity models such as SW-CMM provide organiza-tions a framework for process improvement, helping them iden-tify which process areas need attention to reach a certain maturitylevel. However, many organizations find it difficult to match the

process improvement goals with customer expectations and to predict andmeasure the capability in schedule, effort, and quality. Six Sigma providesthe means to explicitly address these issues, thereby giving organizations

process

� 1995: Reinforced all improvement ini-tiatives under the Tata Business Excel-lence Model umbrella (along the lines ofMalcolm Balridge National QualityAward

� 1996: Adopted SW-CMM to establishdefined and consistently implementedmature software processes throughoutthe organization

� 1998: Applied Six Sigma methods toquantitatively analyze problems withproducts and processes to ensure cus-tomer satisfaction

� 2000: Adopted the People-CMM to in-tegrate people practices to sustain SW-CMM practices and ensure employeesatisfaction3

Currently, 15 of TCS’s 17 developmentcenters in India are operating at SW-CMMLevel 5. Using TCS-QMS, we’ve establishedan infrastructure in these development cen-ters that institutionalizes effective softwareengineering and management processesacross projects.

The TCS-QMS ensures customer satis-faction by

� Designing processes that ensure productand service quality

� Integrating quality control activitiesinto software development

� Stressing quality assurance through de-fect prevention techniques

� Using metrics to manage processes andquality

� Introducing new technologies to ensurecontinuous process improvements

The TCS-QMS process framework hasseveral key features. First, it provides afoundation for effective project manage-ment by tracking cost, schedule, quality,and functionality. We’ve established processdiscipline to repeat earlier successes on proj-ects with similar applications. We also useestimation guidelines, checklists, proce-dures, project management review meet-ings, and automated project-tracking tools4

to satisfy the goals of Level 2 KPAs.Second, TCS-QMS provides an architec-

ture for organization-wide knowledge shar-ing. This architecture includes a dedicatedsoftware process engineering group andprocess owners to institutionalize process

improvements,5 along with well-definedsoftware development life cycles, trainingprograms, tailoring guidelines, service-levelagreements within groups, and product peerreview.6,7 A process assets library fosters insti-tutionalization and sharing of best practices.

Third, TCS-QMS emphasizes data-drivenmanagement of software product and processquality. It does this using different statisticaltechniques and defines targets and tolerancesfor various metrics. It also emphasizes theanalysis of process performance against capa-bility baselines and encourages preventive ac-tions to eliminate process instability.8

Finally, TCS-QMS facilitates continuousprocess improvement with quantitative feed-back from the process itself. It empowersteam members to pilot innovative ideas andtechnologies aimed at process and quality im-provements. We also use statistical tools likePareto analysis and cause-and-effect analysisto identify the source of process problemsand to identify and prioritize improvements.

Blending CMM and Six SigmaThe concepts of SW-CMM Level 4 and

Level 5 and Six Sigma are synergistic. SixSigma is a customer-centric, data-driven ap-proach that focuses on reducing process vari-ation, centering—making the process meancoincide with the process target—and opti-mizing the development process. While SW-CMM provides the basic process infrastruc-ture to systematically apply Six Sigmatechniques,9 Six Sigma helps to build theknowledge and skills necessary to addressSW-CMM’s key requirements, includingquantitative process management, softwarequality management, defect prevention, tech-nology change management, and processchange management. Together, Six Sigmaand CMM help organizations improve mar-ketplace competitiveness and achieve theirbusiness goals.

Process management: The Six Sigma wayUsing the Six Sigma approach has added

many new facets to TCS’s process manage-ment. Based on the Six Sigma principle, wedesigned our metrics program to translatecustomer needs into operational measure-ments. We also carry out detailed analyses ofprocess trends by clearly distinguishing com-mon cause variations from special cause vari-ations. Finally, we eliminate the ambiguities in

M a r c h / A p r i l 2 0 0 3 I E E E S O F T W A R E 4 3

Together, SixSigma and CMM help

organizationsimprove

marketplacecompetitiveness

and achievetheir business

goals.

calculating process capability using the sigmacapability, which considers both process vari-ability and process centering objectives.

Customer-centric metrics program. Our proj-ect teams identify the software process andproduct metrics to track based on QualityFunctional Deployment, a quality tool thatoffers a systematic way to break down cus-tomer needs into manageable actions.Teams typically carry out QFD with cus-tomers, translating customer requirementsinto project requirements. They then rankprocess characteristics in order of impor-tance and identify those that are critical toquality. Table 1 shows a QFD house ofquality matrix, which identifies a project lifecycle’s critical quality characteristics;10 thecolumn totals represent the sum of the prod-ucts of each customer requirement ratingand product requirement.

The team first translates the first column’scustomer requirements into the product re-quirements (object-oriented design, use cases,traceability, and coding standards). The teamthen rates the relationship between productcharacteristics and customer requirementsbased on the relationship’s strength (9 is astrong relationship, 3 a medium relationship,and 1 a weak relationship). In the second

house of quality, the team further analyzesproduct requirements with high technical im-portance to determine the design featuresneeded to meet these requirements.11

Based on the QFD results, the team se-lects in-process metrics that indicateprogress toward the customer-specified im-provement objectives.

The object here is to

� Align measurements with business pri-orities

� Benchmark process and product per-formance against client-established tar-gets, operational requirements, andquality objectives

� Improve the processes to attain these de-fined performance limits

� Facilitate the goal of continuous im-provement

Metrics analysis and management. The con-cept of process variability forms the heart ofprocess metrics analysis.12 Process variationcan be partitioned into two components:

� Natural process variation, frequentlycalled common-cause or system varia-tion, is the naturally occurring fluctua-tion or variation inherent in all processes.

� Special cause variation is typicallycaused by some problem or extraordi-nary occurrence in the system.

We must reduce both the common andspecial cause variation to make the processmore predictable and produce the best out-put. Software organizations typically identifyspecial causes through the presence of pointsoutside the control limits of ± 3σ. TCS pi-loted the Six Sigma approach of analyzingcontrol chart patterns to identify potentialprocess instability indicators. Once we iden-tify an unstable pattern, we find the assigna-ble cause and correct the process. Figure 1shows a process tested for the points outside

4 4 I E E E S O F T W A R E h t t p : / / c o m p u t e r. o r g / s o f t w a r e

Table 1Quality function deployment

First House of Quality Product requirements

Customer requirements Rating Object-oriented design Use cases Traceability Coding standards

Reliability 10 9 9 9 3Maintainability 10 9 3 9 9Portability 7 9 3 3 3Testability 5 3 9 9 1Total of rating × requirement 258 186 246 146

20100

20

10

0

Project number

Rew

ork

inde

x (p

erce

nt) Test 1

Test 1

Test 2Mean = 4.079

1.0 SL = 7.000

2.0 SL = 9.920

3.0 SL = 12.84

–1.0 SL = 1.158

–2.0 SL = –1.762

–3.0 SL = –4.683

Figure 1. Controlchart patterns for a sample process. The process failedtwo tests: Test 1identified points twoand three as outsidethe 3.00 sigma limit(SL); Test 2 identified points15–18 as 1.00 sigmafrom the center line.

the 3σ limit (Test 1) and the points that are 1σfrom the center line (Test 2).

The team analyzed the special causes ofthese patterns and found that the Test 1 fail-ures resulted from tight schedules, whichcreated more rework after internal reviews.They corrected this by loading project teamsmore uniformly. The second set of test re-sults stemmed from team members whoseprior project experience created less rework;the test results are a good indication thatthey shared knowledge across the team.

Process capability calculations. Six Sigma typi-cally monitors the process using controlcharts, which compare control limits withspecification limits to determine process capa-bility. We measure process capability in termsof capability indices Cp, Cpk, and Cpm,12 whichare simplified measures that quickly describethe relationship between process variabilityand the specification limits’ spread. Likemany simplified measures, capability indicesdo not completely describe the process; theyare best used to compare process capability.

In terms of Z capability (sigma capabil-ity), process capability is the number ofstandard deviations that fit between themean and the specification limit; this unitcorresponds to defect probability. We com-pute it as

Z = (x − µ) / σ

where x is the specification limit, µ is themean, and σ is the standard deviation (seeFigure 2).

The essential value of process capabilityanalysis is its ability to predict the percent-age of process characteristics or productsthat the process will produce within specifi-cation limits. Graphically, Z capability isthe distance, measured in standard devia-tions, from the process mean to the specifi-cation limit. In other words, the Z capabil-ity is the area under the normal curve up tothe specification limit.

Tracking process capability in terms of Zensures both our process variation reduc-tion and process centering objectives. It alsohelps us baseline the process capability andclearly compare process performance withspecified targets. Z is a simple metric formeasuring a process’s capability to producedefect-free products.

Figure 3, for example, shows twoprocesses. Process 1 has a mean below thetarget, has large variation, and producesmany defects outside the specification lim-its. In contrast, process 2 is centered on thetarget, has less variation, and has no defectsoutside the specification limits. Comparedwith process 2, process 1 has less Z capabil-ity (as measured by the percentage of nor-mal curve within specification limits). Thus,calculating process capability in terms of Znot only helps project managers understandprocess dynamics but also helps them esti-mate the relative percentage of acceptableand unacceptable characteristics that theprocess is producing on an ongoing basis.

M a r c h / A p r i l 2 0 0 3 I E E E S O F T W A R E 4 5

µ +1σ σ σ+3

z

x+2

Figure 2. Process capability in terms of Z capability: x is thespecification limit, µ is the mean, and σ is the standard deviation.

Lowerspecification

limit

Upperspecification

limit

Target

Process 1

Process 2

DefectDefect

Figure 3. Use of Z capability for predicting process capability.The shaded area denotes defect probability. For a Six Sigmaprocess, the error probability is 0.0000034 (that is, there are only 3.4 defects per million delivered units). For a 3σprocess, the probability of error is 0.0668 (66,800 defects permillion units).

Continuous improvementBlending Six Sigma and SW-CMM helps

us ensure our goal of continuous improve-ments. Six Sigma provides the foundation todefine, measure, analyze, improve, and con-trol the processes,13 whereas methods suchas process mapping and Failure Modes andEffects Analysis (FMEA) help us understandprocess defects and prioritize improvementactions.

As Figure 4 shows, we’ve implementedprocess mapping as a CMM Level 3 practiceto help us understand gaps in an organiza-tion’s standard software process by break-ing the process down into its constituent el-ements. Process mapping shows theordering of process elements and interfaces,and interdependencies among elements. Wealso use process mapping as a CMM Level5 process improvement practice, breakingdown complex processes so that teams cansee them in their entirety. This helps themidentify process delays and loops, identifyproblem spots and improvement targets,and determine the data to collect.

As Table 2 shows, we also use FMEA toidentify improvement areas. Here, it helpsteams

� Identify how a process can fail� Determine failure mode severity� Predict the probability of failure causes

occurring� Find deficiencies in the process control

plan� Prioritize improvement actions

A team assigns potential failure modes aseverity rating from 1 to 10, with 10 havingthe highest impact on customers. It assignsfailure causes a 1 to 10 occurrence rating,with 10 representing the highest likelihood ofoccurrence. Finally, the team assigns a 1 to 10detection rating, with 10 indicating the lowestlikelihood of failure mode or cause detection.

The team then combines these ratings toproduce a risk priority number. It prioritizesthese numbers for the top few problems,identifies improvement actions, assigns re-sponsibilities, and determines time framesfor completion. Once it has implementedthe actions, the team carries out FMEAagain to ensure that it’s reduced the failuremode’s risk.

A more common way to evaluate risk isto assess a risk’s impact and frequency ofoccurrence, then calculate the risk exposureaccordingly. The FMEA approach differs inthat it considers the process controls’ effec-tiveness and helps a team prioritize therisks. Teams can use FMEA effectively forrisk management for SW-CMM Level 3 andfor defect prevention at SW-CMM Level 5.

Benefits to the organizationSix Sigma and SW-CMM complement

each other and together can help an organi-zation meet its process improvement goals.

4 6 I E E E S O F T W A R E h t t p : / / c o m p u t e r. o r g / s o f t w a r e

Process mapping(existing process)

No

YesStudy user

requirements specification

Prepareuse cases

Receive clarifications

Wait for client's reply

Preparesystem

requirements specification

Sendclarifications

to client

Check clarity of

requirements

Figure 4. Process mapping. Implemented as a CMM Level 3process, process mapping breaks down a process so thatteams can identify gaps in standard software practice.

Table 2 Failure mode and effects analysis

Potential failure modes Potential causes Current controls Risk priority number Recommended action

Improper requirements Lack of understanding Review of software 200 Training on software requirementsanalysis of requirements requirements specification management toolsBugs in code Lack of programming skills Code review 180 Coding standardsPoor code review efficiency Untrained reviewers Testing 160 Code review checklistImproper project estimation Lack of knowledge Project management reviews 140 Estimation guidelines

The benefits of blending the methods are il-lustrated by results from a TCS develop-ment center. As Figure 5 shows, the centerused SW-CMM and Six Sigma concepts toreduce its in-process failure cost from 5 to 1percent, thus reducing the cost of quality.

The center also had several process andtechnology improvements for cycle-time re-duction and productivity improvement. Forexample, it implemented

� A statement of work to capture all proj-ect requirements

� The tollgate approach for project execution� Defect prevention checklists� Estimation guidelines � Standard libraries � Guidelines and methodologies� Error-proofing tools

The center’s process capability for prod-uct quality improved from 96 percent in Oc-tober 1999 to 100 percent in September2000. In terms of sigma capability, the cen-ter’s product quality is currently 5.85 (thatis, clients reported no defects in softwareproducts). Process capability for on-time de-livery improved from 2.85 in October 1999to 4.50 in October 2001. Finally, as Figure6 shows, the Six Sigma techniques have re-duced the center’s schedule slippage varia-tion from the 20 to –20 percent range to the2 to –2 percent range. Applying these meth-ods saved the center US$700,000 over athree-year period. We subsequently inte-grated these practices in the TCS QMS anddeployed it across the organization.

W e designed the TCS QMS to en-sure the quality of products andservices and to emphasize contin-

uous improvement. By breaking down thecustomer expectations into process- andproduct-level attributes, the Six Sigma pro-gram offers precise operational definitionsfor improvements. This in turn helps teamstranslate the organization’s strategic goalsinto tactical objectives, which become theprocess improvement initiatives in Six Sigmaprojects. Teams can then feed the derived im-provements into the organization’s CMM-based process framework, which helps defineand institutionalize the improvement actionsinto disciplined and mature processes.

We’ve deployed these blended CMM andSix Sigma concepts in development centers inIndia through the TCS QMS. The next step isto institutionalize the Six Sigma methodologyof Define, Measure, Analyze, Improve, andControl (DMAIC) to all process improve-ment initiatives and to transfer these bestpractices in TCS’s Global Development Cen-tres throughout the world.

References1. M.C. Paulk et al., “Capability Maturity Model Version

1.1,” Software Engineering Institute, Carnegie MellonUniv., CMU/SEI-93-TR-24, DTIC No. ADA263403,Pittsburgh, Pa., 1993.

2. G. Keeni, “The Evolution of Quality Processes at TataConsultancy Services,” IEEE Software, vol. 17, no. 4,July/Aug. 2000, pp. 79–88.

3. G. Keeni, B. Curtis, and C. Kubicki, “Lessons Learnedin Achieving People CMM Level 4 at Tata ConsultancyServices,” Proc. US Software Eng. Process GroupConf., 2002, Software Engineering Institute, CarnegieMellon Univ., 2002; http://seir.sei.cmu.edu/seir/frames/body2.map.P-CMM.html.

4. G. Keeni and D. Biswas, “Evolution of an AutomatedQuality Management System,” Proc. Int’l Symp. FutureSoftware Technology (ISFST 2001), Software Eng. As-soc., Tokyo, 2001, pp. 349–358.

5. G. Keeni, A. Chandra, and S. Dutta, “Structure of theQuality Group for Effective Software Process Improve-ment,” Proc. Software Process Improvement Conf., 2000;www.uni-paderborn.de/cs/ag-schaefer/Tagungen/2000/spi00/spi00.pdf.

6. R. Radice et al., “One on One Inspection,” Proc. Soft-ware Technology Conf., STC Online, Logan, Utah, 2001;www.stc-online.org/stcdocs/stc2001/178/index.htm.

M a r c h / A p r i l 2 0 0 3 I E E E S O F T W A R E 4 7

50250–25

Upper specification limit

Schedule slippage (%)

Lower specification limit

Oct 99Oct 01

Figure 6. Process capability analysis for schedule slippage.

Failu

re c

ost (

perc

ent)

0123456

Oct 9

9

Nov

Dec

Jan

2000 Feb

Mar Apr

May Jun

Jul

Aug

Sep

Oct

Nov

Figure 5. Trend of failure costs. Using SW-CMM and Six Sigmaconcepts, the organization reduced its failure costs to 1 percent.

7. R.P. Nandivada et al., “The 9 Quadrant Model forCode Reviews,” Proc. Asia Pacific Conference on Qual-ity Software, IEEE CS Press, Los Alamitos, Calif., 2000,pp. 188–193.

8. R. Radice, R. Sokhi, and P. Suresh, “The Journey toLevel 4,” Proc. European Software Eng. Process Group(ESEPG 99), Software Technology Transition, 1999;www.stt.com/TCSjourney.PDF.

9. D.N. Card, “Sorting out Six Sigma and CMM,” IEEESoftware, vol. 17, no. 3, May/June 2000, pp. 11–13.

10. R. Prasad and G. Keeni, “Integrated Demanded QualityDeployment and QFD,” Proc. 13th Quality FunctionDeployment Symp., QFD Inst., Ann Arbor, Mich.,2001.

11. M. Murugappan and G. Keeni, “Incorporating Voice ofCustomer in the Six Sigma Way,” SEPG Conf. Tour inAsia-Pacific, 2002; www.softwaredioxide.com/channels/conview.asp?id=6693.

12. S. Sytsma and K. Manley, “Quality Tools Cookbook,”June 1999; www.sytsma.com/tqmtools/tqmtoolmenu.html.

13. M. Murugappan and G. Keeni, “Quality Improvement:The Six Sigma Way,” Proc. Asia Pacific Conf. QualitySoftware, IEEE CS Press, Los Alamitos, Calif., 2000,pp. 248–257.

For more information on this or any other computing topic, please visit ourDigital Library at http://computer.org/publications/dlib.

4 8 I E E E S O F T W A R E h t t p : / / c o m p u t e r. o r g / s o f t w a r e

About the Authors

Mala Murugappan is the Master Black Belt at Global Engineering Development Centrein TCS, Chennai, India. A certified quality analyst, her interests include process and quality im-provement, and Six Sigma quality. She has undergone Six Sigma training for Black Belt andcarried out several Six Sigma projects for quality and productivity Improvement. She has su-pervised more than 50 Six Sigma projects including Green Belt and Black Belt projects, andhas been trained by SEI on Software CMM. She received her BEng in mechanical engineeringfrom the College of Engineering, Guindy, Chennai, and her MS in machine design from the In-dian Institute of Technology, Chennai. Contact her at Global Engineering Development Centre,185-188, Lloyds Road, Chennai 600 086 India; mmala@chennai.tcs.co.in.

Gargi Keeni is vice president at Tata Consultancy Services, where she heads the qualityconsulting practice and assists organizations in achieving their software process improvementgoals. Her research interests include software process improvement and quality managementsystems. She is principal architect of the Integrated Project Management System, which pro-vides all the necessary operational metrics from the project level to the organizational level.She has more than 20 years’ experience in software project management, software tools de-velopment, and software process improvement. Under her leadership as Corporate QualityHead, TCS’s 15 development centers were assessed at Software CMM Level 5, and TCS was thefirst organization to be assessed at Level 4 of the People CMM v2.0. She is an authorized Soft-ware CMM Lead Assessor, People CMM Lead Assessor, Lead Appraiser for SCAMPI, an examiner for the Tata Business Excel-lence Model, and a Certified Quality Analyst. She has a PhD in nuclear physics from Tohoku University, Sendai, Japan. Sheis a member of the Computer Society of India and IEEE. Contact her at Tata Consultancy Services, SDF building, Salt LakeElectronics Complex, Sec-V, Block EP and GP, Kolkata 700 091, India; gargi@ieee.org.

IEEE Pervasive Computing delivers the latest peer-revieweddevelopments in pervasive, mobile, and ubiquitous computing and acts as a catalyst for realizing the vision of pervasive (orubiquitous) computing, described by Mark Weiser nearly adecade ago. In 2003, look for articles on

• Security & Privacy

• The Human Experience

• Building Systems That Deal with

Uncertainty

• Interfacing with the Physical World

SUBSCRIBE NOW!

http://computer.org/pervasive

M. SatyanarayananCarnegie Mellon Univ. and Intel Research Pittsburgh

Associate EICs

IEEE Pervasive Computing

Editor in Chief

Roy Want, Intel Research; Tim Kindberg, HP Labs; Deborah Estrin, UCLA; Gregory Abowd, GeorgiaTech.;

Nigel Davies, Lancaster University and Arizona University