Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Copyright 2005 Northrop Grumman Corporation
A Practical Guideto ImplementingLevels 4 and 5CMMI Technology Conference & User Group14-17 November 2005
Rick Hefner, Ph.D.Northrop Grumman Corporation
Copyright 2005 Northrop Grumman Corporation2
Agenda
� An Overview of Levels 4 and 5� New Behaviors� Benefits – Project, Organizational, Customer� Making the Business Case
� Understanding the CMMI Process Areas� Organizational Process Performance� Quantitative Project Management� Causal Analysis & Resolution� Organizational Innovation & Deployment
� Strategies for Adoption� Lessons Learned� Links to Six Sigma
SM SCAMPI, SCAMPI Lead Appraiser, and SEI are service marks of Carnegie Mellon University.® Capability Maturity Model Integration and CMMI are registered in the U.S. Patent & Trademark Office.
Copyright 2005 Northrop Grumman Corporation3
Causal Analysis and ResolutionOrganizational Innovation and Deployment5 Optimizing
4 QuantitativelyManaged
3 Defined
2 Managed
Quantitative Project ManagementOrganizational Process Performance
Requirements DevelopmentTechnical SolutionProduct IntegrationVerificationValidationOrganizational Process FocusOrganizational Process DefinitionOrganizational TrainingRisk ManagementIntegrated Project Management (for IPPD*)Integrated Teaming*Integrated Supplier Management**Decision Analysis and ResolutionOrganizational Environment for Integration*
Requirements ManagementProject PlanningProject Monitoring and ControlSupplier Agreement ManagementMeasurement and AnalysisProcess and Product Quality AssuranceConfiguration Management
1 Performed
Process AreasLevel
Proactivemanagement
PredictiveManagement
Reactive mgmt.(plan, track, and
correct)
Management Styles in the CMMI
Qualitativeimprovement
Quantitativeimprovement
Project Organizational
Copyright 2005 Northrop Grumman Corporation4
Exercise –What is Managing by Variation?
� Suppose your project conductedseveral peer reviews of similarcode, and analyzed the results� Mean = 7.8 defects/KSLOC� +3σ = 11.60 defects/KSLOC� -3σ = 4.001 defects/KSLOC
� What would you expectthe next peer review toproduce in terms ofdefects/KSLOC?
� What would you think ifa review resulted in 11defects/KSLOC?
� 3 defects/KSLOC?
151050
12
11
10
9
8
7
6
5
4
3
Observation Number
Indi
vidu
alV
alue
I Chart for Defects
Mean=7.8
UCL=11.60
LCL=4.001
Copyright 2005 Northrop Grumman Corporation5
Exercise –What is Required to Manage by Variation?
� What is needed to develop thestatistical characterization of aprocess?
� The process has to bestable (predictable)� Process must be
consistently performed� Complex process may
need to be stratified(separated into simplerprocesses)
� There has to be enoughdata points to statisticallycharacterize the process� Processes must occur
frequently within asimilar context (projector organization)
� May need to stratify thedata
151050
12
11
10
9
8
7
6
5
4
3
Observation Number
Indi
vidu
alV
alue
I Chart for Defects
Mean=7.8
UCL=11.60
LCL=4.001
Copyright 2005 Northrop Grumman Corporation6
Freq
uen
cy
Hours to Close
10
Process Example: Closing a change request
1 4 7 10 13 16
15
20
5
0
What Is a Data Distribution?
� The data distributiongives a frequency ofoccurrence of eachvalue in a data set
� Once we characterizethe data distribution, wecan predict futurevalues or assign aprobability to anyspecific value
� Data distributions mayrepresent population orprocess data
� When applied toprocess data - assumesstatistical stability
A bell curve isa Normal orGaussianDistribution
Frequencydistributionof hours toclose
Copyright 2005 Northrop Grumman Corporation7
The Normal Distribution
0
50
100
150
200
250
300
350
-16 -12 -8 -4 0 4 8 12 16 20
Data Value
Frequency
of O
ccure
nce
� Commonly called a bell curve;formally called a Gaussiandistribution
� Applies to continuous data (e.g.time, weight, distance)
� Characteristic shape:� Mean, mode, and median are
all the same value� Standard deviation is
independent of the mean
� Normal distributions are not alaw, but they are often observedin natural data� Often the distribution of
difference between plan andactual dates or budget versusactual
Copyright 2005 Northrop Grumman Corporation8
Characteristics of a Normal Curve
Once we understand whatkind of distribution ourdata comes from, we canpredict future values orassign a probability toany specific point
• Probability x will occur
• Probability a value < x
• Probability a value > x
• Probability a value willoccur between x and y
De
nsi
ty
3210-1-2-3
0.4
0.3
0.2
0.1
0.0
34%34%
14%14%
2%2%
99.7%
95.4%
68.3%
�=meanσ=1 standarddeviation
Copyright 2005 Northrop Grumman Corporation9
The Rayleigh Distribution…
0
50
100
150
200
250
300
350
0 1 2 3 4 5 6 7 8
Data Value
Frequency
of O
ccure
nce
� A skewed distribution
� Used to model effort or defectdensity in software developmentwhere the abscissa representsdevelopment phase if they areroughly the same duration
� Rayleigh distribution is a specialcase of the Weibull distribution
Copyright 2005 Northrop Grumman Corporation10
What Is a Control Chart?
� A time-ordered plot of processdata points with a centerlinebased on the average andcontrol limits that bound theexpected range of variation
� Control charts are one of themost useful quantitative toolsfor understanding variation
Copyright 2005 Northrop Grumman Corporation11
What Are the Key Features of a Control Chart?
ID
Def
ects
per
Lin
eof
Code
252321191715131197531
0.10
0.09
0.08
0.07
0.06
_U=0.07825
UCL=0.09633
LCL=0.06017
Tests performed with unequal sample sizes
UpperControl
LimitProcess“Average”
LowerControl
Limit
Timeorderedx-axis
Individualdata points
U Chart of Defect Detected in Requirements Definition
Copyright 2005 Northrop Grumman Corporation12
What is Special Cause and Common CauseVariation?
151050
12
11
10
9
8
7
6
5
4
3
Observation Number
Indi
vidu
alV
alue
I Chart for Defects
Mean=7.8
UCL=11.60
LCL=4.001
� Common Cause Variation� Routine variation that comes
from within the process� Caused by the natural variation
in the process� Predictable (stable) within a
range
� Special Cause Variation� Assignable variation that comes
from outside the process� Caused by a unexpected
variation in the process� Unpredictable
151050
15
10
5
0
Observation Number
Indi
vidu
al V
alue
I Chart for Defects_
1
Mean=7.467
UCL=13.36
LCL=1.578
Copyright 2005 Northrop Grumman Corporation13
What Is a Stable Process?
ID
Def
ects
per
Line
ofCo
de
252321191715131197531
0.10
0.09
0.08
0.07
0.06
_U=0.07825
UCL=0.09633
LCL=0.06017
Tests performed with unequal sample sizes
U Chart of Defects Detected in Requirements Definition
All data points withinthe control limits. Nosignals of specialcause variation.
Copyright 2005 Northrop Grumman Corporation14
Hearing Voices
� Voice of the process = the natural bounds of processperformance
� Voice of the customer = the goals established for the product andprocess performance
stable process
Capable process = +
product conformance
� Process capability may be determined for the� Organization� Product line� Project� Individual
� Typically, the higher the level of analysis, the greater the variation
Copyright 2005 Northrop Grumman Corporation15
What if the Process Isn’t Stable?
� You may be able to explainout of limit points byobserving that they are dueto an variation in theprocess� E.g., peer review held on
Friday afternoon� You can eliminate the
points from the data, ifthey are not part of theprocess you are trying topredict
� You may be able to stratifythe data by an attribute ofthe process or attribute ofthe corresponding workproduct� E.g., different styles of
peer reviews, peerreviews of different typesof work products
0
5
10
15
20
25
1 11 21 31 41 51 61 71
UCL
_X
Def
ects
per
com
pone
nt
Component #
0
5
10
15
20
25
1 11 21 31 41 51 61 71
UCL
_X
Def
ects
per
com
pone
nt
Component #
Out of limit points
0
5
10
15
20
25
1 11 21 31 41 51 61 71
UCL
_X
Def
ects
per
com
pone
nt
Component #
0
5
10
15
20
25
1 11 21 31 41 51 61 71
UCL
_X
Def
ects
per
com
pone
nt
Component #
0
5
10
15
20
25
1 11 21 31 41 51 61 71
UCL
_X
Def
ects
per
com
pone
nt
Component #
0
5
10
15
20
25
1 11 21 31 41 51 61 71
UCL
_X
Def
ects
per
com
pone
nt
Component #
Copyright 2005 Northrop Grumman Corporation16
Common Challenges for Engineering
� Data are often discrete ratherthan continuous, e.g., defects
� Observations often are scarce
� Processes are aperiodic
� Size of the the object oftenvaries, e.g., software module
� Data distributions may not benormal
Copyright 2005 Northrop Grumman Corporation17
How Do I Address These Challenges?
� Employ control chart types thatspecifically deal with discrete datadistributions, e.g., u-charts and p-charts
� Use control charts that compensatefor widely variable areas ofopportunity
� Transform non-normal continuousdata to normal data beforeconstructing a control chart (BlackBelt topic)
� Cross check control charts withhypothesis tests (later module) wherefew data points exist
Copyright 2005 Northrop Grumman Corporation18
Typical Choices in Industry
� Most customers care about:� Delivered defects� Cost and schedule
� So organizations try topredict:� Defects found throughout
the lifecycle� Effectiveness of peer
reviews, testing� Cost achieved/actual
(Cost Performance Index –CPI)
� Schedule achieved/actual(Schedule PerformanceIndex – SPI)
Defect Detection Profile
0.00
20.00
40.00
60.00
80.00
100.00
120.00
140.00
160.00
180.00
Req'mts Design Code Unit Test Integrate Sys Test Del 90 Days
Phase
Def
ects
/KSL
OC
All ProjectsNew Process
Process performance• Process measures (e.g., effectiveness, efficiency, speed)• Product measures (e.g., quality, defect density).
Copyright 2005 Northrop Grumman Corporation19
How Can High Maturity Help?
0 5 10 15
0
5
10
15
Observation Number
Indi
vidu
al V
alue
Mean=7.268
UCL=11.17
LCL=3.363
1 2
� By measuring both the meanand variation, the project/organization can assess the fullimpact of an “improvement”
� Can focus on reducing thevariation (making the processmore predictable)� Train people on the process� Create procedures/checklists� Strengthen process audits
� Can focus on increasing themean (e.g., increaseeffectiveness, efficiency, etc.)� Train people� Create checklists� Reduce waste and re-work� Replicate best practices from
other projects
� Can do both
processshift
Copyright 2005 Northrop Grumman Corporation20
What Can a Level 4 Organization/Project Do?
� Both can:� Determine whether processes are behaving consistently or have stable
trends (i.e., are predictable)� Eliminate special causes of variation� Identify processes that show unusual (e.g., unpredictable) behavior� Identify the implementation of a process which performs best
� The organization can:� Characterize the process performance of the organization’s standard
process� Develop models to predict process performance
� The organization’s projects can:� Use this data to decide how to tailor the organization’s standard process� Statistically manage selected subprocesses to achieve project quality and
process objectives
Would your organization/projects find this valuable?
Copyright 2005 Northrop Grumman Corporation21
Level 5
Level 3
What Does Level 5 Add?
Organizational Process Focus� Goals are qualitative (e.g., get
better)� The effects of the improvements
are not estimated or measured
Organizational Innovation &Deployment
� Goals are quantitative (e.g.,reduce variation by X%, reducemean by Y%)
� Incremental improvements� Innovative improvements - cause
a major shift in processcapability
� Potential improvements areanalyzed to estimate costs andimpacts (benefits)
� Improvements are piloted toensure success
� Improvements are measured interms of variation and mean
Causal Analysis
� Both the project andorganization can betterdetermine cause and effectrelationships
� This can be used to focusprocess improvements andpreventive actions
Builds onLevel 4
capabilities
Copyright 2005 Northrop Grumman Corporation22
How Does Level 4 & 5 Benefit the Customer?
� The project fixes the source ofdefects to prevent future defects� Causal analysis
� The project benefits fromimprovements found and provenon other projects
� Organizational innovationand deployment
� Problem behaviors arerecognized faster, enablingquicker resolution
� Quantitative projectmanagement
� More accurate estimates� Organizational processperformance
Better Products and Services Produced Faster And CheaperBetter Products and Services Produced Faster And CheaperBetter Products and Services Produced Faster And Cheaper
“How Does How Does High Maturity Benefit the Customer”, R. Hefner,Systems & Software Technology Conference,. 2005
Copyright 2005 Northrop Grumman Corporation23
Agenda
� An Overview of Levels 4 and 5� New Behaviors� Benefits – Project, Organizational, Customer� Making the Business Case
� Understanding the CMMI Process Areas� Organizational Process Performance� Quantitative Project Management� Causal Analysis & Resolution� Organizational Innovation & Deployment
� Strategies for Adoption� Lessons Learned� Links to Six Sigma
Copyright 2005 Northrop Grumman Corporation24
The Project Manager’s Dilemma at Level 3
I want to use the organization’sstandard process, but…
… Does it’s performance and qualitymeet my customer’s expectations?
… If not, how should I tailor theprocess?
Copyright 2005 Northrop Grumman Corporation25
� Characterize the performance of the organization’s standardprocess statistically
� Develop models to help a project manager determine theperformance they would be likely to get by using the standardorganizational process model, given their project’s characteristics
What Should the Organization Do to Help theProject Manager?
customer andproject
objectives
OrganizationProject
Standard Process Baselines & ModelsHistorical Data
Tailored ProcessProject Results
Copyright 2005 Northrop Grumman Corporation26
Organizational Process Performance
SG 1 Establish Performance Baselines and ModelsBaselines and models that characterize theexpected process performance of the organization'sset of standard processes are established andmaintained.
SP 1.1 Select ProcessesSelect the processes or process elements in theorganization's set of standard processes that are tobe included in the organization's processperformance analyses.
SP 1.2 Establish Process Performance MeasuresEstablish and maintain definitions of the measuresthat are to be included in the organization's processperformance analyses.
SP 1.3 Establish Quality and Process-PerformanceObjectivesEstablish and maintain quantitative objectives forquality and process performance for theorganization.
SP 1.4 Establish Process Performance BaselinesEstablish and maintain the organization's processperformance baselines.
SP 1.5 Establish Process Performance ModelsEstablish and maintain the process performancemodels for the organization's set of standardprocesses.
Selected subprocesses, NOT thewhole process
Objectives deal with eliminatingsources of variation, not setting“stretch” goalsThe organization meets thesegoals by modifying the standardprocess, not driving the projects
Baselines characterize the “voice of theprocess”, based on the existinghistorical data
• What is the current mean andvariation?
May need to subgroup the data
Models allow projects to estimatetheir quantitative performance basedon the historical data of otherprojects executing the process
Copyright 2005 Northrop Grumman Corporation27
Quantitative Baselines and Models
� From this baseline (and model), what would you predict the nextpeer review to produce in terms of defects/KSLOC?
� What other models could be developed to help predict?
151050
12
11
10
9
8
7
6
5
4
3
Observation Number
Indi
vidu
alV
alue
I Chart for Defects
Mean=7.8
UCL=11.60
LCL=4.001
Copyright 2005 Northrop Grumman Corporation28
BU
ILD N
Process Architecture
RequirementsAnalysis
ArchitecturalDesign
DetailedDesign Fabrication Integration &
Test
ProcessTailoring
ProjectPlanning
ProjectTracking
RiskManagement
SupplierManagement
QualityAssurance
ConfigurationManagement
DataManagement
Measurement &Analysis
DecisionAnalysis &Resolution
Integration
Peer R
eview
Systems
Testing
Systems
Validation
Fagan Inspection
Walkthrough
Desk C
heck
Process ElementsAlternative Process Elements
Copyright 2005 Northrop Grumman Corporation29
The Project Manager’s Challenge at Level 4
I understand the capabilities of the organization’s standardprocess, but…
… What are the project’s quality and process performanceobjectives?
… How should I tailor the process?
… What project subprocesses doI need to quantitatively manage?
Copyright 2005 Northrop Grumman Corporation30
Quantitative Project Management (Goal 1)
SG 1 Quantitatively Manage the ProjectThe project is quantitatively managed usingquality and process-performance objectives.
SP 1.1 Establish the Project’s ObjectivesEstablish and maintain the project’s qualityand process performance objectives.
SP 1.2 Compose the Defined ProcessSelect the subprocesses that compose theproject’s defined process based on historicalstability and capability data.
SP 1.3 Select the Subprocesses that Will BeStatistically ManagedSelect the subprocesses of the project'sdefined process that will be statisticallymanaged.
SP 1.4 Manage Project PerformanceMonitor the project to determine whether theproject’s objectives for quality and processperformance will be satisfied, and identifycorrective action as appropriate.
Quality: defect levels of keywork products or deliverables
Process: productivity, efficiency,effectiveness of the project’sprocesses
Rationale for how the project tailoredthe organization’s standard process,in order to meet their quality &process performance objectives
• E.g., adding procedures to reducevariation
Assumes the standard processincludes subprocesses to select from
Only some subprocesses selected forstatistical management
• Need not be the same as thoseselected by the organization, or otherprojects
Monitoring against the objectivesestablished in SP 1.1
Copyright 2005 Northrop Grumman Corporation31
Outer Loop
SP 1.1 Establish the Project’s ObjectivesEstablish and maintain the project’s qualityand process performance objectives.
SP 1.4 Manage Project PerformanceMonitor the project to determine whetherthe project’s objectives for quality andprocess performance will be satisfied, andidentify corrective action as appropriate.
SP 1.3 Select the Subprocesses that WillBe Statistically ManagedSelect the subprocesses of the project'sdefined process that will be statisticallymanaged
Copyright 2005 Northrop Grumman Corporation32
Tailoring
RequirementsAnalysis
ArchitecturalDesign
DetailedDesign Fabrication Integration &
Test
ProcessTailoring
ProjectPlanning
ProjectTracking
RiskManagement
SupplierManagement
QualityAssurance
ConfigurationManagement
DataManagement
Measurement &Analysis DAR
Integration
Peer R
eview
Systems
Testing
Systems
Validation
Fagan Inspection
Walkthrough
Desk C
heckRequirements
AnalysisArchitectural
DesignDetailedDesign Fabrication Integration &
Test
ProcessTailoring
ProjectPlanning
ProjectTracking
RiskManagement
SupplierManagement
QualityAssurance
ConfigurationManagement
DataManagement
Measurement &Analysis DAR
Organization
Project
SP 1.2 Compose the Defined ProcessSelect the subprocesses that compose the
project’s defined process based onhistorical stability and capabilitydata.
Copyright 2005 Northrop Grumman Corporation33
Selecting Subprocesses to be Statistically Managed
� Which processes do you need to be stable(predictable) in order to achieve your project’sobjectives?� For these, eliminate special causes, characterize
the process, and predicatively manage
� The time needed to perform this practice is longand often unpredictable� Many processes can not be made predictable
� Example – objectives for delivered defects� Defect detection (peer review, unit testing, system
testing)� Defect insertion (requirement definition,
architecture, design, integration)
SP 1.3 Select the Subprocesses that WillBe Statistically ManagedSelect the subprocesses of the project'sdefined process that will be statisticallymanaged
Copyright 2005 Northrop Grumman Corporation34
Quantitative Project Management (Goal 2)
SG 2 Statistically Manage Subprocess PerformanceThe performance of selected subprocesseswithin the project's defined process isstatistically managed.
SP 2.1 Select Measures and Analytic TechniquesSelect the measures and analytic techniquesto be used in statistically managing theselected subprocesses.
SP 2.2 Apply Statistical Methods to UnderstandVariationEstablish and maintain an understanding ofthe variation of the selected subprocessesusing the selected measures and analytictechniques.
SP 2.3 Monitor Performance of the SelectedSubprocessesMonitor the performance of the selectedsubprocesses to determine their capability tosatisfy their quality and processperformance objectives, and identifycorrective action as necessary.
SP 2.4 Record Statistical Management DataRecord statistical and quality managementdata in the organization’s measurementrepository.
Type of analysis to be performed(e.g., control charts)
Key is understanding variationin the selected subprocesses(e.g., be able to computestandard deviation), NOT justmetrics
Given the stability and variation inthe subprocesses, will we be ableto meet our project-level qualityand process performanceobjectives?
This data is used to help selectsubprocesses in tailoring (SP 1.2)
Copyright 2005 Northrop Grumman Corporation35
Inner Loop
SP 1.1 Establish the Project’s ObjectivesEstablish and maintain the project’s qualityand process performance objectives.
SP 1.4 Manage Project PerformanceMonitor the project to determine whetherthe project’s objectives for quality andprocess performance will be satisfied, andidentify corrective action as appropriate.
SP 1.3 Select the Subprocesses that WillBe Statistically ManagedSelect the subprocesses of the project'sdefined process that will be statisticallymanaged
SP 2.2 Apply Statistical Methods toUnderstand VariationEstablish and maintain an understanding ofthe variation of the selected subprocessesusing the selected measures and analytictechniques.
SP 2.3 Monitor Performance of theSelected SubprocessesMonitor the performance of the selectedsubprocesses to determine their capabilityto satisfy their quality and processperformance objectives, and identifycorrective action as necessary.
Copyright 2005 Northrop Grumman Corporation36
Sample Measures for Quantitative Analysis
� Time between failures� Critical resource utilization� Number and severity of defects in the
released product� Number and severity of customer
complaints concerning the pro-videdservice
� Number of defects removed byproduct verification activities(per-haps by type of verification,such as peer reviews and testing)
� Defect escape rates� Number and density of defects by
severity found during the first yearfollowing product delivery or start ofservice
� Cycle time� Amount of rework time� Requirements volatility (i.e., number
of requirements changes per phase)� Ratios of estimated to measured
values of the planning parameters(e.g., size, cost, and schedule)
� Coverage and efficiency of peerreviews (i.e., number/amount ofproducts reviewed compared to total,number of defects found per hour)
� Test coverage and efficiency (i.e.,number/amount of products testedcompared to total, number of defectsfound per hour)
� Effectiveness of training (i.e., percentof planned training completed andtest scores)
� Reliability (i.e., mean time-to-failureusually measured during inte-grationand systems test)
� Percentage of the total defectsinserted or found in the differentphases of the project life cycle
� Percentage of the total effortexpended in the different phases ofthe project life cycle
� Profile of subprocesses understatistical management (i.e., numberplanned to be under statisticalmanagement, number currently beingstatistically managed, and numberthat are statistically stable)
� Number of special causes of variationidentified
Source: Interpreting the CMMI, Margaret Kulpa and Kent Johnson
Copyright 2005 Northrop Grumman Corporation37
New Questions at Level 4
� What characteristics of the organizational standardprocess would be useful to understand?
� Which subprocesses would be useful to understand,for predictive purposes?
� Are these subprocesses predictable (stabilizable)?
� What data should the organization collect?
� To what level of detail should the organizationalstandard process go?
� What differences in project subprocesses arepermissible? How do they impact the historical data?
Copyright 2005 Northrop Grumman Corporation38
What Does Level 5 Add to the Project/Organization?
� Casual Analysis & Resolution� Identify and analyze causes of defects and other problems� Take specific actions to remove the causes
� The project may take actions to prevent the occurrence of thosetypes of defects and problems in the future
� Most projects use CAR-type methods and tools at Level 4, toidentify and eliminate special cause variations, i.e., to stabilizethe processes
� At Level 5, CAR is used to eliminate common cause variation
effects
Copyright 2005 Northrop Grumman Corporation39
Causal Analysis & Resolution
SG 1 Determine Causes of DefectsRoot causes of defects and other problemsare systematically determined.
SP 1.1 Select Defect Data for AnalysisSelect the defects and other problems foranalysis.
SP 1.2 Analyze CausesPerform causal analysis of selected defectsand other problems and propose actions toaddress them.
Can apply to any cause and effectrelationship, not just defectsTypically, projects will establish alist of potential areas in which toapply CAR, and select some fromthat list
Determine cause and effect(e.g., fishbone diagram,brainstorming) and potentialimprovement action listSelect some actions on the listto implement
SG 2 Address Causes of DefectsRoot causes of defects and other problemsare systematically addressed to prevent theirfuture occurrence.
SP 2.1 Implement the Action ProposalsImplement the selected action proposals thatwere developed in causal analysis.
SP 2.2 Evaluate the Effect of ChangesEvaluate the effect of changes on processperformance.
SP 2.3 Record DataRecord causal analysis and resolution datafor use across the project and organization.
Implemented for each selected action• Evidence will provide samples
Measures the effect of thechange
The effect need not be positive
Copyright 2005 Northrop Grumman Corporation40
Organizational Innovation and Deployment
SG 1 Select ImprovementsProcess and technologyimprovements that contribute tomeeting quality and process-performance objectives are selected.
SP 1.1 Collect and Analyze ImprovementProposalsCollect and analyze process- andtechnology-improvement proposals.
SP 1.2 Identify and Analyze InnovationsIdentify and analyze innovativeimprovements that could increasethe organization’s quality andprocess performance.
SP 1.3 Pilot ImprovementsPilot process and technologyimprovements to select which onesto implement.
SP 1.4 Select Improvements forDeploymentSelect process- and technology-improvement proposals fordeployment across the organization.
SG 2 Deploy ImprovementsMeasurable improvements to theorganization's processes andtechnologies are continually andsystematically deployed.
SP 2.1 Plan the DeploymentEstablish and maintain the plans fordeploying the selected process andtechnology improvements.
SP 2.2 Manage the DeploymentManage the deployment of theselected process and technologyimprovements.
SP 2.3 Measure Improvement EffectsMeasure the effects of the deployedprocess and technologyimprovements.
The effects should be measuredquantitatively (as opposed toqualitatively in OPF)
Copyright 2005 Northrop Grumman Corporation41
Agenda
� An Overview of Levels 4 and 5� New Behaviors� Benefits – Project, Organizational, Customer� Making the Business Case
� Understanding the CMMI Process Areas� Organizational Process Performance� Quantitative Project Management� Causal Analysis & Resolution� Organizational Innovation & Deployment
� Strategies for Adoption� Lessons Learned� Links to Six Sigma
Copyright 2005 Northrop Grumman Corporation42
Achieving Levels 4 and 5 is Less Predictable
� The time it takes to achieveLevels 2 and 3 is driven by theresources available� Learn/implement new project
practices� Create organizational assets -
policies, processes, training,etc.
� Practices are first performed;effectiveness is improved overtime
� The time it takes to achieveLevels 4 and 5 is driven by theability to stabilize processes� Choosing processes that can
be stabilized� Establishing the right metrics
and methods� Collecting enough data� Effectiveness is required to
perform the practices
Causal Analysis and ResolutionOrganizational Innovation and Deployment5 Optimizing
4 QuantitativelyManaged
3 Defined
2 Managed
Quantitative Project ManagementOrganizational Process Performance
Requirements DevelopmentTechnical SolutionProduct IntegrationVerificationValidationOrganizational Process FocusOrganizational Process DefinitionOrganizational TrainingRisk ManagementIntegrated Project Management (for IPPD*)Integrated Teaming*Integrated Supplier Management**Decision Analysis and ResolutionOrganizational Environment for Integration*
Requirements ManagementProject PlanningProject Monitoring and ControlSupplier Agreement ManagementMeasurement and AnalysisProcess and Product Quality AssuranceConfiguration Management
1 Performed
Process AreasLevel
Copyright 2005 Northrop Grumman Corporation43
Lessons Learned
Based on 20 Northrop Grumman CMMI Level 5 organizations� Level 3 metrics, measurement processes, and goal setting are generally
inadequate for Levels 4 and 5� Better definitions of the measures� Lower level metrics of lower level subprocesses� Stratifying the data properly
� When operating at Level 3, it is difficult to predict the measurementimprovements needed� Trying to understand and stabilize the key subprocesses will naturally
drive you to the right metrics
� Projects have different quality and process performance needs, andshould select different subprocesses to quantitatively manage� This will also slow adoption, and complicate the organizational baselines
and models
� Six Sigma is an enabler for higher maturity� Focus on data, measurement systems, process improvement� Tying improvements to business goals� Tools and methods support the level 4/5 analysis tasks
Copyright 2005 Northrop Grumman Corporation44
What is Six Sigma?
� Six Sigma is a management philosophy based on meetingbusiness objectives by striving for perfection� A disciplined, data-driven methodology for decision making and
process improvement
� Six Sigma consists of several integrated methods:� Process Management� Voice of the Customer� Change Management� Tools for Measuring Variation and Change� Business Metrics
� Leading-edge companies are applying Six Sigma to engineeringwork
Copyright 2005 Northrop Grumman Corporation45
Importance of Reducing Variation
� To increase process performance, you have to decrease variation
� Less variation means� Greater predictability in the process� Less waste and rework, which lowers costs� Products and services that perform better and last longer� Happier customers
Defects Defects
Too early Too late
Delivery Time
Reducevariation
Delivery Time
Too early Too late
Spread of variationtoo wide compared
to specifications
Spread of variationnarrow compared to
specifications
Copyright 2005 Northrop Grumman Corporation46
A Typical Six Sigma Project in Engineering
� Customers express concern that software defects are causingfrequent failures in the field
� A Six Sigma team is formed to scope the problem, collect data,and determine the root cause
� The team’s analysis of the data determines that poorlyunderstood interface requirements account for 90% of theproblems in the field
� The interface problems are corrected
� The organization’s requirements solicitation process is modifiedto ensure future projects do not encounter similar problems
Copyright 2005 Northrop Grumman Corporation47
DMAIC Process Steps
DEFINE Set project goals and objectives
MEASURE Narrow range of potential causes and establishbaseline capability level
ANALYZE Evaluate data/information for trends, patterns, causalrelationships and "root causes“
IMPROVE Develop, implement and evaluate solutions targetedat identified root causes
CONTROL Make sure problem stays fixed and new methods canbe further improved over time
Copyright 2005 Northrop Grumman Corporation48
How Six Sigma Helps Level 4-5 Organizations
� Six Sigma provides specificmethods and tools for:� Quantitative process
management of Level 4� Causal Analysis and
Resolution of Level 5
� Six Sigma projects provide amechanism for selecting andimplementing improvements� Addresses Organizational
Innovation and Deployment� Can extend beyond Level 5
Copyright 2005 Northrop Grumman Corporation49
References
� “Squeezing Variation for Profit, “ Don Corpron, CMMI Technology Conference and UserGroup, 2005
� “Using a Process Database to Facilitate Transition to Level 4”, Rick Hefner, InternationalConference on Applications of Software Measurement, 2002.
� Purcell, Leitha, “Experiences Using Six Sigma in a SW-CMM® Based Process ImprovementProgram”, American Society for Quality Six Sigma Conference, January 2001.
� Interpreting the CMMI: A Process Improvement Approach, Margaret Kulpa and Kent A.Johnson, Auerbach Publications, 2003
� Card, David, “Sorting Out Six Sigma and the CMM”, IEEE Software, May 2000.
� iSixSigma, http://www.isigsima.com