sixsigma and cmmi everything you wanted to know, but didn’t know who to ask
DESCRIPTION
SixSigma and CMMI Everything You Wanted to Know, But Didn’t Know Who to Ask. Tom Lienhard Six Sigma BlackBelt. Today’s Mission. I have 90 Minutes. BlackBelt training is 160 hrs. Intro to CMMI is 24 hrs. QUICK Introduction to Six Sigma Relationship of Six Sigma to the CMMI - PowerPoint PPT PresentationTRANSCRIPT
Strive to be
Level 5
1
SixSigma and CMMIEverything You Wanted to Know,
But Didn’t Know Who to Ask
Tom Lienhard
Six Sigma BlackBelt
Strive to be
Level 5
2
I have 90 Minutes
Today’s Mission
• QUICK Introduction to Six Sigma
• Relationship of Six Sigma to the CMMI
• Applying Six Sigma
• Examples of Six Sigma Tool Usage to Improve
BlackBelt training is 160 hrs Intro to CMMI is 24 hrs
Strive to be
Level 5
3
What Do You Think Six Sigma Is All About?
Benchmark?
Method?
Metric?
Vision?Philosophy?
Stretch Goal?
Pipe Dream? Disease?
Strive to be
Level 5
4
What Is Six Sigma?
A customer-driven management methodology to significantly
increase customer satisfaction through reducing and eliminatingvariation and thus defects.
A concept developed by Motorola to achieve a desired level of quality, virtually defect-free products and processes
With this definition, six sigma is applicable to any development or manufacturing process
Strive to be
Level 5
5
The Many Facets of Sigma
Sigma is a letter inthe Greek alphabet “Sigma” is used to designate the
distribution about the mean of any process or product characteristic(standard deviation)
“Sigma value” indicates how often defects are
likely to occur
“Sigma level” is the number of standard deviations between the process center and the CLOSEST spec limit
Strive to be
Level 5
6
Premise for Six Sigma
• Sources of variation can be• Identified• Quantified• Mitigated by control or prevention
• The elements of a process responsible for variation are called the “6 M’s”• Man (people)• Machine• Material• Method• Measurement• Mother Nature
Strive to be
Level 5
7
Variation is the Enemy
• Variation exists in everything
• If it is assumed that everything is a result of some process, then it is safe to conclude that the process introduces product variation
• Variation in the product is due to variation in the process
• Variation comes in two forms - common and special cause
We Must Slay the Enemy
Strive to be
Level 5
8
Common and Special CausesCommon Causes• Continuously active in the process• Predictable
Special Cause• Extraordinary events• Can assign a reason• Can do something about it
87654321
11
10
9
8
7
6
5
4
3
2
X=7.4
3.0SL=9.2.
-3.0SL=5..6
Special causes
Common causes
Control limits reflect the built in or natural process variationControl limits are not related to standards or specificationThe control limits allow predictions about the future to be made
Strive to be
Level 5
9
0
2000
4000
6000
8000
10000
12000
14000
6/1
4/9
9
6/2
8/9
9
7/1
2/9
9
7/2
6/9
9
8/9
/99
8/2
3/9
9
9/6
/99
9/2
0/9
9
10
/4/9
9
10
/18
/99
11
/1/9
9
11
/15
/99
11
/29
/99
12
/13
/99
12
/27
/99
1/1
0/0
0
1/2
4/0
0
2/7
/00
2/2
1/0
0
3/6
/00
3/2
0/0
0
We
ekl
y S
tan
dar
d H
ou
rs
Demonstrated
UCL
2 SIGMA
1 SIGMA
MEAN
1 SIGMA
2 SIGMA
LCL
Planned
Historical output to stores in standard hours
99.7% probability that output of an “in-control”process will fall between the upper control limit(UCL) and the lower control limit (LCL).
68% probability that output of an “in-control”process will fall between the 1 sigma limits.
95% probability that output of an “in-control”process will fall between the 2 sigma limits.
Planned requirements in standard hours
Average historical output
1
2
3
4
5
8
6
7
1 One point more than three sigmas from the mean
Tests for out-of-control conditions in a control chart
2 Nine points in a row on the same side of the mean
3 Six points in a row all increasing or decreasing
4 Fourteen points in a row, alternating up and down
5
6 Four out of five points more than 1 sigma from the mean
7 Fifteen points in a row within 1 sigma of the mean
8 Eight points in a row more than 1 sigma from the mean
Two out of three points more than 2 sigmas from the mean
Not shownControl limits are calculated based on the week-to-week variation of the demonstrated output
The control limits are used as a predictor of future output if the historical output is in control (i.e. does not fail any of the 8 tests below) and the process has not changed significantly
Violation of any of the 8 tests below by the plannedrequirements indicates a nearly zero probability thatthe actual output can meet the plan. A majorprocess change or re-plan is required.
Shows the difference between planned and actual for a given week
Reading a Control Chart
Strive to be
Level 5
10
largest source of variation (components of variation, analysis of variance)
factors which have the greatest influence on variation (Process Map, FMEA, Screening Design of Experiments)
where to set factors to reduce variation and maximize the output (Optimizing Design of Experiments)
Six Sigma is All About Variation
Six Sigma tools help identify
types of variation (control chart) 6s
87654321
110
100
90
80
70
60
50
40
30
20
1
1
1
X=74.46
3.0SL=92.36
-3.0SL=56.55
Special causes
Common causes
Strive to be
Level 5
11
Six Sigma Tool Box – Example Usage
• Thought Process Map• Process Map• Failure Modes and Effect Analysis• Pareto Chart• Control Chart• Confidence Intervals• Product Scorecard• Postmortem• Measurement System Evaluation• Regression Analysis• Design of Experiments• Main Effects Plot• Interaction Plot
6s
Strive to be
Level 5
12
Identify variation
Reduce or eliminate variation
Maximize output and deliver less defects
Satisfied Customer
Customer Satisfaction Through Reducing Defects
Characterize variation
Strive to be
Level 5
13
Six Sigma is Based on Statistics
Mean (Process Center)
1 Standard Deviation
Strive to be
Level 5
14
Process CenterThe distribution of Our Process output
REWORKREWORK GOOD GOOD
Upper SpecificationLimit
Lower SpecificationLimit
Determined by the customer
Determined by the customer
Incapable Process
• We have REWORK because some of what we make is waste
• Process is WIDER than the specification limits
Capable Process
• We make as much as the customer needs and have very little waste
• Process FITS within the specification limits
States of a Process
Process Center Upper SpecificationLimit
Lower SpecificationLimit
Determined by the customer
Determined by the customer
GOOD GOOD
Strive to be
Level 5
15
PCB
PCB : Process Capability Baseline = “normal” variation for the sample
If you take the bell shaped curve and turn it on its side, you have a control chart
X
X : Sample Mean = arithmetic average of a set of values (process center)
s
: s Standard Deviation = the square root of the variance
UCL
UCL : Upper Control Limit = +3s from the Sample Mean
LCL
LCL : Lower Control Limit = -3s from the Sample Mean
X Bar What? Standard Who?
Strive to be
Level 5
16
Four Possible States of a Process
Spec. limit
Control LimitCapable but
Unstable
Brink of Chaos
Unstable andnot Capable
Chaos
Stable butnot Capable
Threshold
Stable andCapable
Ideal
Strive to be
Level 5
17
So Long As We Satisfy the Requirements, Right?
Pilot A Pilot B
Which pilot would you want to fly with?
Both within spec, same average landing
Why?
Strive to be
Level 5
18
LowerSpec Upper
Spec
LowerSpec
UpperSpec
Target Target
ProcessCenter
ProcessCenter
4s
s s ss
6s
s s s ss
1 StandardDeviation
s
1 StandardDeviation
Pilot A is at 4 sigmaIn order to be six sigma, like Pilot B, the
variation in Pilot A’s landings must be reduced
Moving Towards Six Sigma
Pilot A’s Distribution Pilot B’s Distribution
Strive to be
Level 5
19
Pilot A and B at Sky Harbor Airport
750
463
100
9
0.3
0.005
0 100 200 300 400 500 600 700 800
1
2
3
4
5
6
Sigma Level
Bad Landings / Day
0.005
Pilot A
Pilot B
Wouldn’t stay in business too long...
Strive to be
Level 5
20
Higher Sigma Equates toMajor Reduction in Defects
500,000
308,500
66,810
6,210 233 3.40
50,000
100,000
150,000
200,000
250,000
300,000
350,000
400,000
450,000
500,000
1 2 3 4 5 6
Sigma Level
Def
ects
/ M
illi
on
Op
po
rtu
nit
ies
Strive to be
Level 5
21
Six Sigma -The Other Story
1s - 170 misspelled words per page in a book
2s - 25 misspelled words per page
3s - 1.5 misspelled words per page
4s - 1 misspelled word per 30 pages
5s - 1 misspelled word in a set of encyclopedias
6s - 1 misspelled word in all the books in a library
Understand you customers’ needs And balance with cost
Strive to be
Level 5
22
CMMI Meets Six Sigma
Strive to be
Level 5
23
CMMI Reminder
Optimizing
Managed
Defined
QuantitativelyManaged
Level Predicted Performance
Time/$/Quality/...
Time/$/Quality/...
Time/$/Quality/...
Time/$/Quality/...
the best practices reside in the projects
the organization has a defined common process which includes the collection of product and process metrics
all the process assets accumulated from Level 2 and 3 are used to quantitatively understand the project
the organization concentrates on improving the process to gain an advantage in quality, productivity, and timeliness
Strive to be
Level 5
24
Level 2 - Measure the Project
The initial focus for measurement activities is at the project level. As the organization moves toward Level 3, the measures prove useful for addressing organization and/or enterprise-wide information needs.
Measurement and Analysis (and GP 2.8) Monitor and control the [PA Name] process against the plan for performing the process and take corrective action
Strive to be
Level 5
25
Level 3 – Understand the Organizational Practices
At Level 3, project data is collected and analyzed by the organization. Thresholds are set based on good engineering judgment which trigger corrective action
Level 2/3 set the stage for quantitative management (statistical process control).• Repeatable, Consistent Process• Adequate Resources• Defined, Consistent Process Measures• Data Collected From Above Process• Assurance of Process Compliance
Strive to be
Level 5
26
Level 4 - Stabilizing the ProcessAt maturity level 4, an organization applies quantitative and statistical techniques to manage process performance and product quality. Quality and process performance is understood in statistical terms. Special causes of process variation are identified and, where appropriate, the sources of special causes are corrected to prevent future occurrences.
87654321
110
100
90
80
70
60
50
40
30
20
1
1
1
X=74.46
3.0SL=92.36
-3.0SL=56.55
Special causes
Design Peer Review
CodePeer Review
TestPeer Review
Formal TestPeer Review
Tota
l # o
f D
efec
ts
Life Cycle Phase
Customer Before Delivery
Req’ts Peer Review
Strive to be
Level 5
27
Level 4 - Stabilizing the Process
When performance falls outside normal range of process performance• Identify the reason
• Take corrective action when appropriate
A critical distinction between maturity level 3 and 4 is the predictability of process performance. At maturity level 4, the
performance of processes is controlled using statistical and other quantitative techniques, and is quantitatively predictable.
Strive to be
Level 5
28
Level 5 - Optimizing the Process
At maturity level 5, an organization continually improves its processes based on a quantitative understanding of the common causes of variation inherent in processes.
When performance is within normal range of process performance but unable to meet the established objectives• Narrow the performance limits• Shift the performance limits
87654321
150
100
50
0
X=74.46
3.0SL=136.0
-3.0SL=12.91
Common causes
Strive to be
Level 5
29
Am I Acting 4 or 5?
A critical distinction between levels 4 and 5 is the type of process variation addressed.
• At maturity level 4, the organization is concerned with addressing special causes of process variation and providing statistical predictability of the results. Although processes may produce predictable results, the results may be insufficient to achieve the established objectives.
• At maturity level 5, the organization is concerned with
addressing common causes of process variation and changing the process (to shift the mean of the process performance or reduce the inherent process variation experienced) to improve process performance and to achieve the established quantitative process-improvement objectives
Strive to be
Level 5
30
Six Sigma allows the organization to statistically analyze whether a process change was actually an improvement to the normal range of process performance
Level 5-Statistical Process Improvement
Level 5 involves continuous improvement to the processes with the goal of improving quality, productivity and decreasing cycle time
Old Process New Process
Cyc
le T
ime
Strive to be
Level 5
31
Six Sigma Tools Enable Levels 4&5
The Six Sigma methodology provides the tools that implement
• Quantitative Project Management,• Organization Process Performance• Causal Analysis and Resolution • Organizational Performance Management
A Level 5 organization is one that has implemented, among other things, statistical process control (SPC)
6s
Strive to be
Level 5
32
Applying Six Sigma
Develop Infrastructure
Quantitatively UnderstandProcess Performance
ContinuousMeasurable
Improvement
PROCESS
Strive to be
Level 5
33
Develop Infrastructure
• Commitment to Process Improvement• Repeatable, Consistent Process• Adequate Resources• Defined, Consistent Process Measures• Data Collected From Above Process• Assurance of Process Compliance
Only required for the window you are looking through
Strive to be
Level 5
34
Quantitatively Understand Process Performance
Use Six Sigma tools to identify sources of variation (quantitatively understand process performance) and then act to eliminate or reduce those sources
Measurements of processperformance on project
Quantitatively understand the process performance
Discover sources of variation
Determine the magnitude of the effects due to process changes
Implement improvements
Monitor the impact
Strive to be
Level 5
35
Six Sigma – Real Life Example
1-11-11-11-11-1
80
60
4080
60
4080
60
4080
60
4080
60
40
Program
Experience
Training
Criteria
Num People
1
-1
1
-1
1
-1
1
-1
1
-1
Interaction Plot (data means) for % Match
Determine the magnitude
8765432Subgroup 1
1101009080706050403020
Sam
ple
Mea
n
1
1
1
X=74.46
3.0SL=92.36
-3.0SL=56.55
Monitor the impact
PlanningCustomerRqmts. Analysis
DesignImplementation
Test Formal Test
Customer BeforeTOTALLeaked
Planning2.03 0 0 0 0 0 0 0 2.03 0Customer0 1.8 1.4 0 4.06 0 16.15 0 23.41 21.61Rqmts. Analysis0 0 7.32 12.04 32.77 41.6 56.09 0.79 150.61143.29Design 0 0 0.13 41.99 8.2 23.2 118.94 5.28 197.74155.75
Implementation 0 0 0.17 0.5 154 90.3 88.88 23.3 357.15203.15Test 0 0 0 0.16 0.03 19.92 4.5 0 24.61 4.69
Formal Test 0 0 0 0 0 2.34 149.25 0 151.59 2.34
Customer Before 0 0 0 0 0 0 2.7 13.6 16.3 13.6
TOTAL2.03 1.8 9.02 54.69 199.06177.36436.5142.97923.44544.43
Measurements
8765432Subgroup 1
1101009080706050403020
Sam
ple
Mea
n
1
1
1
X=74.46
3.0SL=92.36
-3.0SL=56.55
Statistically understand
Strive to be
Level 5
36
Continuous Measurable Improvement
Measurements of processperformance of project
Statistically understand the process capability
Discover sources of variation
Determine the magnitude of theeffects due to process changes
Implement improvementsMonitor the impact
Measurements of processperformance of project
Statistically understand the process capability
Discover sources of variation
Determine the magnitude of theeffects due to process changes
Implement improvementsMonitor the impact
• Can be next biggest hitter• Can be next quickest• Can be next logical problem• Can be anything the org. determines
Strive to be
Level 5
37
Six Sigma Tool Box – Example Usage
• Thought Process Map• Process Map• Failure Modes and Effect Analysis• Pareto Chart• Control Chart• Confidence Intervals• Product Scorecard• Postmortem• Measurement System Evaluation• Regression Analysis• Design of Experiments• Main Effects Plot• Interaction Plot
6s
Strive to be
Level 5
38
Thought Process Map
Problem to be solved
What is theprocess?
Process Mapcurrent process
What are thepossible
weaknesses?
FMEA/CEPostmortems
What is the currentdata ?
Gather data What data?
What about older projects?No, data not available
Data from
New projectsOrganize data to
effectively analyze
What data?What want to know?Estimate rework
by phase
How do wemeasure theprocess?
Look for differentrequirementsDifferent customersMultiple perspective
Involve multiple projectsMultiple disciplines
BARRIER-Lack of data?
Red - Question or expected result Blue - Answer or actual result
BARRIER-Lack of data?
Is there one?How Determine?
What arex's, y's?
Howtell?
Data forFMEA
Underlined - Barrier
How well is itworking now?
What is datashowing us?
Refine estimate as site databecomes available
Drill down into thedata/NEM/Control
Charts
Is the data anygood?
Validate MeasurementSystem MSE(KAPPA/ICC
or Nested Design)
DOE
Update process toensure data has
higher confidencerate/ train
NotAdequate
What areimportant factors?
BARRIER-People's time
Determine actionBased on data Make and
communicateimprovements
Did change causeimprovement?
Set up controlplan and Use
Control Charts toMonitor
Effectiveness ofImprovements
Adequate
NewslettersLiaison Meetings
BARRIER-Projects not
required to follow?
BARRIER-People's time
BARRIER-People's time
Choice of factors
Question to answer
Stable process
Yes
Identify andRemove Special
Causes
No
CharacterizeOptimizeProcess
Strive to be
Level 5
39
Process Map
Activity/Sub-process
C - InputN - InputS - InputX - Input
OutputOutputOutput
Rework
Defects?
Rework
X - Critical (statistical proven critical)N - Noise (can't or choose not to control)S - SOP (the standard way to do it)C - Controllable (can be changed to see effect)
Activity/ProcessInputs Outputs
Rework
50,000 ft level
500 ft level
Activity/Sub-process
Activity/Sub-process
Activity/Sub-process
Activity/Sub-process
Rework
Defects?
No No
C - InputN - InputS - InputC - Input
X - InputN - InputS - InputX - Input
C - InputS - InputS - InputS - Input
C - InputN - InputS - InputX - Input
OutputOutput
Output
OutputOutput
OutputOutputOutput
Strive to be
Level 5
40
Process or Product Name:
In-Phase Peer ReviewPrepared by: Tom Lienhard & Team
Page ____ of ____
Responsible:Tucson Software Engineering Process Group
FMEA Date (Orig) 23 Sept 99 (Rev) 29 Sept 99
Process Step/Input
Potential Failure Mode Potential Failure EffectsSEV
Potential CausesOCC
Current ControlsDET
RPN
Actions Recommended
What is the process step/ Input under
investigation?
In what ways does the process step go wrong?
What is the impact on the Key Output Variables (Customer Requirements) or internal requirements?
What causes the process step to go wrong?
What are the existing controls and procedures (inspection and test) that prevent eith the cause or the Failure Mode? Should include an SOP number.
What are the actions for reducing the
occurrance of the Cause, or improving detection? Should
have actions only on high RPN's or easy
fixes.
Peer Review Planning
No Review Team identified upfront w/review package(roles & responsibilities)
Product not reviewed by appropriate disciplines
6
SW plans do not require this
1
SEPG/SQA and peer review of plans
1 6
None
6
Lack of process awareness
6
Moderator to ensure review package complete (contains reviewers)
6 216
Re-train moderator and conduct process evaluations
No product evaluation criteria identified with review package
Product not reviewed to customer and/or process requirements
9
SW plans do not require this
1
SEPG/SQA and peer review of plans 1 9
None
9
Lack of process awareness
6
Moderator to ensure review package complete (contains review criteria)
6 324
Re-train moderator and conduct process evaluations
No notice or agenda with review package
Team not able to give adequate review time 6
SW plans do not require this1
SEPG/SQA and peer review of plans 1 6
None
6
Lack of process awareness
6
Moderator to ensure review package complete (contains notice and agenda)
6 216
Re-train moderator and conduct process evaluations
Product Review Product not reviewed Defects not found
9
Adequate time not given to review
6
Moderator and SQD ensure adequate time was given to review product 3 162
None - cultural thing
9
No or inadequate evaluation criteria
6
Moderator to ensure review package complete (contains evaluation criteria)
9 486
Create generic checklists for site (to highest level) and conduct process evaluations
9
Inappropriate reviewers
6
Moderator to ensure appropriate reviewers
9 486
Update command media to identify required participants on notice & agenda and conduct process evluations
Metrics not captured Organization quantitative data incorrect/incomplete 3
SW plans do not require this1
SEPG/SQA and peer review of plans 1 3
None
3
Lack of process awareness
9
Moderator to ensure review process followed (metrics captured)
6 162
Re-train moderator and conduct process evaluations
Process/Product Failure Modes and Effects Analysis
(FMEA)
1 3
Severity9 - Defects go to customer6 - Defects cause rework3 - Data not collected1 - No harm/no foul
Occurrence:9 - Regular occurrence6 - Occurs more than occasionally3 - Occurs occasionally1 - Rare occurrence
Detection:9 - Nothing in place6 - Based on individuals3 - Check in place, usually works1 - Check in place & working
FMEA
Strive to be
Level 5
41
Potential causes (factors) which kept showing up over and over on the FMEA:• Inappropriate review team (“wrong” moderator, dominant, inexperienced, or
yes-people made up the team)• Lack of process awareness (both unintentional and deliberate)• No or inadequate review criteria (review what is there not what is missing,
biased review based on experience with phase)
Plan to minimize the occurrence and increase the detection :• Update the process to highlight required participants, their roles and
responsibilities on the Notice and Agenda• Roll-out Peer Review training• Perform peer review process evaluations• Generate common evaluation criteria for all work products that can be used
across the entire organization
Use what was learned about factors as an input into DOE:
What Was Learned
Strive to be
Level 5
42
Planning Customer Rqmts.
Analysis
Design Implement
ation
Test Formal
Test
Customer
Before TOTAL LeakedPlanning
2.03 0 0 0 0 0 0 0 2.03 0Customer
0 1.8 1.4 0 4.06 0 16.15 0 23.41 21.61Rqmts.
Analysis 0 0 7.32 12.04 32.77 41.6 56.09 0.79 150.61 143.29Design
0 0 0.13 41.99 8.2 23.2 118.94 5.28 197.74 155.75Implement
ation 0 0 0.17 0.5 154 90.3 88.88 23.3 357.15 203.15Test
0 0 0 0.16 0.03 19.92 4.5 0 24.61 4.69Formal
Test 0 0 0 0 0 2.34 149.25 0 151.59 2.34Customer
Before 0 0 0 0 0 0 2.7 13.6 16.3 13.6
TOTAL 2.03 1.8 9.02 54.69 199.06 177.36 436.51 42.97 923.44 544.43
Phase Detected
Ph
ase
Intr
odu
ced
Data has been changed to protect the innocent
Cost of Rework due to defects-in days
Strive to be
Level 5
43
$0
$50,000
$100,000
$150,000
$200,000
$250,000
$300,000
$350,000
$400,000
Plannin
g
Custom
er
Reqs A
nalys
is
Design
Implem
enta
tion
Test
Formal T
est
Custom
er B
efor
e
Phase Detected
Cu
mu
lati
ve C
ost
s
Existing Detection
Improved Detection
Existing defect detection profile
Goal defect detection profile
Same number of total defects introduced in the same phases
IMPACT ONBOTTOM LINE
Cost Savings Goal
Strive to be
Level 5
44
100
90
80
70
60
50
40
30
20
Per
cent
Def
e cts
Fou
nd I
n P
hase
Goal
Plan
ning
Cu s
t . Pl
an
Req
’ts
Des
i gn
Imp l
em.
Test
Form
al T
est
Cu s
t om
er
These 3 phases accountfor > 92% of rework
Distribution Characteristics of Data
Strive to be
Level 5
45
2520151050
1.0
0.5
0.0
Sample Number
Pro
portio
n
P Chart for In Phase
P=0.7678
3.0SL=1.000
-3.0SL=0.00E+00
P Chart – Determines Stability
But we sampled only 3 projects - what does the population look like?
Strive to be
Level 5
46
25 26 27 28 29 30 31
95% Confidence Interval for Mu
20 25 30 35
95% Confidence Interval for Median
Variable: Req
A-Squared:P-Value:
MeanStDevVarianceSkewnessKurtosisN
Minimum1st QuartileMedian3rd QuartileMaximum
20.6813
1.6737
25.0000
0.3380.200
28.6667 3.214610.3333
-1.54539
3
25.000025.000030.000031.000031.0000
36.6521
20.2026
31.0000
Anderson-Darling Normality Test
95% Confidence Interval for Mu
95% Confidence Interval for Sigma
95% Confidence Interval for Median
Descriptive Statistics
60 70 80 90 100
95% Confidence Interval for Mu
0 20 40 60 80 100 120 140
95% Confidence Interval for Median
Variable: Cust1
A-Squared:P-Value:
MeanStDevVarianceSkewnessKurtosisN
Minimum1st QuartileMedian3rd QuartileMaximum
7.512
13.237
55.000
0.4460.081
70.666725.4231646.3331.72000
3
55.000 55.000 57.000100.000100.000
133.821
159.777
100.000
Anderson-Darling Normality Test
95% Confidence Interval for Mu
95% Confidence Interval for Sigma
95% Confidence Interval for Median
Descriptive Statistics
55 60 65 70 75
95% Confidence Interval for Mu
40 50 60 70 80 90
95% Confidence Interval for Median
Variable: Design
A-Squared:P-Value:
MeanStDevVarianceSkewnessKurtosisN
Minimum1st QuartileMedian3rd QuartileMaximum
42.2518
4.8377
55.0000
0.2590.384
65.3333 9.291686.3333
-1.18512
3
55.000055.000068.000073.000073.0000
88.4149
58.3951
73.0000
Anderson-Darling Normality Test
95% Confidence Interval for Mu
95% Confidence Interval for Sigma
95% Confidence Interval for Median
Descriptive Statistics
65 67 69 71 73
95% Confidence Interval for Mu
60 70 80
95% Confidence Interval for Median
Variable: Implem
A-Squared:P-Value:
MeanStDevVarianceSkewnessKurtosisN
Minimum1st QuartileMedian3rd QuartileMaximum
58.6271
2.1042
65.0000
0.2120.536
68.6667 4.041516.3333
0.722109
3
65.000065.000068.000073.000073.0000
78.7062
25.3995
73.0000
Anderson-Darling Normality Test
95% Confidence Interval for Mu
95% Confidence Interval for Sigma
95% Confidence Interval for Median
Descriptive Statistics
82 84 86 88 90 92
95% Confidence Interval for Mu
72 82 92 102
95% Confidence Interval for Median
Variable: test
A-Squared:P-Value:
MeanStDevVarianceSkewnessKurtosisN
Minimum1st QuartileMedian3rd QuartileMaximum
73.985
2.868
82.000
0.1930.616
87.6667 5.507630.3333-2.7E-01
3
82.000 82.000 88.000 93.000 93.000
101.348
34.614
93.000
Anderson-Darling Normality Test
95% Confidence Interval for Mu
95% Confidence Interval for Sigma
95% Confidence Interval for Median
Descriptive Statistics
97 98 99
95% Confidence Interval for Mu
96 97 98 99 100 101
95% Confidence Interval for Median
Variable: Form Tst
A-Squared:P-Value:
MeanStDevVarianceSkewnessKurtosisN
Minimum1st QuartileMedian3rd QuartileMaximum
95.516
0.521
97.000
0.1890.631
98 110
3
97.000 97.000 98.000 99.000 99.000
100.484
6.285
99.000
Anderson-Darling Normality Test
95% Confidence Interval for Mu
95% Confidence Interval for Sigma
95% Confidence Interval for Median
Descriptive Statistics
Confidence Intervals
Strive to be
Level 5
47
phase
pe
rce
nt
Planning Cust. Reqs Design Implem Test FormalTst
Cust.
40
90
140
0
If I were a betting man…. the true population means are within these intervals
7
134
21
3742
88
59
78
74
101
96
100
14
139
100
Goal
…Gives Us Population Range
Strive to be
Level 5
48
Analysis of Variance for percent
Source DF SS MS F Pphase 7 10841.9583 1548.8512 8.701 0.000project 16 2848.0000 178.0000Total 23 13689.9583
Variance Components
Source Var Comp. % of Total StDevphase 456.950 71.97 21.376project 178.000 28.03 13.342Total 634.950 25.198
Phase variationProject variation
ANOVA Tells Us the Variation is Between Subgroups
Strive to be
Level 5
49
If no action is taken 95% confident that• the Requirements Phase will find between 21% - 37% of defects in phase• the Design Phase will find between 42% - 88% of defects in phase• the Implementation Phase will find 59% - 78% of defects in phase
Variation between the phases (72%) is greater than variation between projects (28%)• need to work largest source of variation - what changed between, what didn’t, etc.
80 %World Class
100 %
SW Phase
Green -customer processRed - < 92 5 of reworkBlue - testing process
What Was Learned
Strive to be
Level 5
50
StdOrder RunOrder Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 % Match1 1 -1 1 -1 -1 -1 705 2 -1 -1 -1 -1 1 553 3 -1 -1 -1 1 -1 307 4 -1 1 -1 1 1 602 5 -1 -1 1 -1 -1 706 6 -1 1 1 -1 1 804 7 -1 1 1 1 -1 658 8 -1 -1 1 1 1 609 9 1 -1 -1 -1 -1 4513 10 1 1 -1 -1 1 5515 11 1 -1 -1 1 1 3511 12 1 1 -1 1 -1 5514 13 1 -1 1 -1 1 8010 14 1 1 1 -1 -1 8012 15 1 -1 1 1 -1 5516 16 1 1 1 1 1 70
Fractional Factorial DesignFactors: 5 Base Design: 5, 16 Resolution: VRuns: 16 Replicates: 1 Fraction: 1/2Blocks: none Center pts (total): 0
Design of Experiment (DOE)
Strive to be
Level 5
51
20100-10
1
0
-1
Effect
No
rmal
Sco
re
C
A
B
Normal Probability Plot of the Effects(response is % Match, Alpha = .10)
A: ExperienB: TrainingC: CriteriaD: Num Peop
151050
B
C
A
AB
AD
ACD
AC
D
CD
BD
BCD
BC
ABD
ABC
Pareto Chart of the Effects(response is % Match, Alpha = .10)
A: ExperienB: TrainingC: CriteriaD: Num Peop
Shows Training, Criteria, Experienceas the influential factors
Data looks pretty normal
Strive to be
Level 5
52
Fractional Factorial FitEstimated Effects and Coefficients for % (coded units)
Term Effect CoefConstant 60.312 Program -1.875 -0.938 Experien 13.125 6.562 Training 19.375 9.687 Criteria -13.125 -6.562 Num Peop 3.125 1.562 Program*Experien -1.875 -0.937 Program*Training 4.375 2.187 Program*Criteria 1.875 0.937 Program*Num Peop -1.875 -0.937 Experien*Training -5.625 -2.812 Experien*Criteria 4.375 2.188 Experien*Num Peop -4.375 -2.187 Training*Criteria -1.875 -0.938 Training*Num Peop 1.875 0.937 Criteria*Num Peop 1.875 0.937
Analysis of Variance for % (coded units)
Source DF Seq SS Adj SS Adj MS F PMain Effects 5 2932.8 2932.8 586.56 * *2-Way Interactions 10 440.6 440.6 44.06 * *Residual Error 0 0.0 0.0 0.00Total 15 3373.4
Shows same thingTraining, Criteria, Experience
as the influential factors
Strive to be
Level 5
53
1-1 1-1 1-1 1- 1 1- 1
70
65
60
55
50
% M
atc h
Main Effects Plot (data means) for % Match
Process pretty robust tothese factors
Factor 1 Factor 2 Factor 3 Factor 4 Factor 5
Want to set these factors “high”
DOE Main Effects Plot
Strive to be
Level 5
54
1-11-11-11-11-1
80
60
4080
60
4080
60
4080
60
4080
60
40
1
-1
1
-1
1
-1
1
-1
1
-1Factor 1
Factor 2
Factor 3
Factor 4
Factor 5
DOE Interaction Plot
Strive to be
Level 5
55
Experience is a no brainer - more is better
Training is no brainier - have to have it!
Number of people was eye opening - didn’t make too much difference Since more people cost more money, keep it at 2
Criteria was a shock
Need to do follow-up since I was not present to take copious notes
Follow-up revealed that having criteria limited the scope of the review,reviewed only what was there, did not use criteria as intended
What Was Learned
Strive to be
Level 5
56
Strive to be
Level 5
57
AFTER IMPROVEMENTS
PHASE DETECTED
PHASEPlanning Customer Rqmts.
Analysis
Design Implementa
tion
Test Formal
Test
Customer
Before TOTAL
IPlanning
235 2 0 3 0 0 0 0 240
NCustomer
0 22 22 7 4 0 0 0 55
TRqmts.
Analysis 0 1 161 13 7 2 0 0 184
RDesign
0 0 3 173 111 2 0 0 289
OImplementa
tion 0 0 0 2 342 20 0 1 365
DTest
0 0 0 2 2 22 1 0 27
UFormal
Test 0 0 0 0 0 0 2 1 3
CCustomer
Before 0 0 0 0 0 0 0 3 3
D TOTAL 235 25 186 200 466 46 3 5 1166
Planning Customer Rqmts. Analysis
Design Implementation
Test Formal Test
Customer Before TOTAL
Planning29 0 0 0 0 0 0 0 29
Customer0 12 2 0 2 0 5 0 21
Rqmts. Analysis 0 0 61 14 29 26 71 1 202Design
0 0 1 323 82 29 38 2 475Implement
ation 0 0 1 5 220 43 44 10 323Test
0 0 0 2 1 249 30 0 282Formal
Test 0 0 0 0 0 13 597 0 610Customer Before 0 0 0 0 0 0 1 4 5
TOTAL 29 12 65 344 334 360 786 17 1947
Number of defects identified by phase introduced/phase detected (from Product Scorecard)
Number of defects identified by phase introduced/phase detected (from Product Scorecard)
Strive to be
Level 5
58
AFTER IMPROVEMENTS
PHASE DETECTED
PHASEPlanning Customer Rqmts.
Analysis
Design Implementa
tion
Test Formal
Test
Customer
Before TOTAL
IPlanning
98% 1% 0% 1% 0% 0% 0% 0% 21%
NCustomer
0% 40% 40% 13% 7% 0% 0% 0% 5%
TRqmts.
Analysis 0% 1% 88% 7% 4% 1% 0% 0% 16%
RDesign
0% 0% 1% 60% 38% 1% 0% 0% 25%
OImplementa
tion 0% 0% 0% 1% 94% 5% 0% 0% 31%
DTest
0% 0% 0% 7% 7% 81% 4% 0% 2%
UFormal
Test 0% 0% 0% 0% 0% 0% 67% 33% 0%
CCustomer
Before 0% 0% 0% 0% 0% 0% 0% 100% 0%
D TOTAL 20% 2% 16% 17% 40% 4% 0% 0% 100%
0%
0%10%
Planning Customer Rqmts.
Analysis
Design Implement
ation
Test Formal
Test
Customer
Before TOTALPlanning
100% 0% 0% 0% 0% 0% 0% 1%Customer
0% 57% 0% 10% 0% 24% 1%Rqmts.
Analysis 0% 0% 30% 7% 14% 13% 35% 0% 10%Design
0% 0% 0% 68% 17% 6% 8% 0% 24%Implement
ation 0% 0% 0% 2% 68% 13% 14% 3% 17%Test
0% 0% 0% 1% 0% 88% 11% 0% 14%Formal
Test 0% 0% 0% 0% 0% 2% 98% 0% 31%Customer
Before 0% 0% 0% 0% 0% 0% 20% 80% 0%
TOTAL 1% 1% 3% 18% 17% 18% 40% 1% 100%
Percentage of defects identified by phase introduced/phase detected (from Product Scorecard)
Percentage of defects identified by phase introduced/phase detected (from Product Scorecard)
Strive to be
Level 5
59
PHASE DETECTED
PHASEPlanningCustomerRqmts.
AnalysisDesignImplementa
tionTest Formal
TestCustomer
Before TOTAL Leaked
IPlanning
16.45 2 0 0.51 0 0 0 0 18.96 2.51
NCustomer
0 3.3 15.4 7 8.12 0 0 0 33.82 30.52
TRqmts. Analysis 0 1 19.32 11.18 7.91 3.2 0 0 42.61 23.29
RDesign
0 0 0.39 22.49 11.1 1.6 0 0 35.58 13.09
OImplementa
tion 0 0 0 0.2 239.4 42 0 2.33 283.93 44.53
DTest
0 0 0 0.16 0.06 1.76 0.15 0 2.13 0.37
UFormal Test 0 0 0 0 0 0 0.5 0.58 1.08 0.58
CCustomer
Before 0 0 0 0 0 0 0 10.2 10.2 10.2
D TOTAL 16.45 6.3 35.11 41.54 266.59 48.56 0.65 13.11 428.31 125.09
Planning Customer Rqmts. Analysis
Design Implementation
Test Formal Test
Customer Before TOTAL Leaked
Planning2.03 0 0 0 0 0 0 0 2.03 0
Customer0 1.8 1.4 0 4.06 0 16.15 0 23.41 21.61
Rqmts. Analysis 0 0 7.32 12.04 32.77 41.6 56.09 0.79 150.61 143.29Design
0 0 0.13 41.99 8.2 23.2 118.94 5.28 197.74 155.75Implement
ation 0 0 0.17 0.5 154 90.3 88.88 23.3 357.15 203.15Test
0 0 0 0.16 0.03 19.92 4.5 0 24.61 4.69Formal
Test 0 0 0 0 0 2.34 149.25 0 151.59 2.34Customer
Before 0 0 0 0 0 0 2.7 13.6 16.3 13.6
TOTAL 2.03 1.8 9.02 54.69 199.06177.36436.51 42.97 923.44 544.43
AFTER IMPROVEMENTS
Cost of Rework due to Defects in Days (from Product Scorecard)
Cost of Rework due to Defects in Days (from Product Scorecard)
Strive to be
Level 5
60
AFTER IMPROVEMENTS
87654321
1 0 0
9 0
8 0
7 0
6 0
5 0
4 0
3 0
2 0
C 4 4
Ba
se
line
Three Phases
87654321
1 0 0
9 0
8 0
7 0
6 0
5 0
4 0
3 0
2 0
C 5 1
Imp
ro
ve
d Three Phases
95% Confident That Two of he ThreePhase Will be World-Class. The ThirdHad Drastic Improvement
95% Confident That Two of he ThreePhase Will be World-Class. The ThirdHad Drastic Improvement
What happened here? Remember weare measuring % not #. We went from282 to 27, a huge improvement in rework $ and effort, but % wise it was a decline
Strive to be
Level 5
61
AFTER IMPROVEMENTS
Rework Effort in Dollars by Phase
$0
$50,000
$100,000
$150,000
$200,000
$250,000
Plannin
g
Custom
er
Rqmts. A
nalys
is
Design
Implem
enta
tion
Test
Formal T
est
Custom
er B
efor
e
TOTAL
Software Development Phase
Do
llar
s
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Leaked Rework
In-Phase Rework
% Leaked
Cost of Rework due to DefectsCost of Rework due to Defects
Rework Effort in Dollars by Phase
$0
$50,000
$100,000
$150,000
$200,000
$250,000
Software Development Phase
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Leaked Rework
In-Phase Rework
% Leaked
Strive to be
Level 5
62
$0
$50,000
$100,000
$150,000
$200,000
$250,000
$300,000
$350,000
$400,000
Planning Customer Reqs Analysis Design Implementation Test Formal Test CustomerBefore
Phase Detected
Existing Detection
Goal
Actual
Baseline defect cost profile
Improved defect cost profile
Goal defect cost profile
Bottom Line SavingsBottom Line Savings
$120,000 in six months on just 3 software projects
(Notice the increased up front spending)
Strive to be
Level 5
63
AFTER IMPROVEMENTS
Three Phase Where Project Concentrated(Three Phases With > 92% of Rework)Three Phase Where Project Concentrated(Three Phases With > 92% of Rework)
Recalculating the control limitsto see if there is significance….
654321
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
Sample Number
Pro
port
ion
P Chart for In Phase
P=0.5751
3.0SL=0.6845
-3.0SL=0.4658
654321
1.0
0.9
0.8
0.7
0.6
0.5
0.4
Sample Number
Pro
port
ion
P Chart for In Phase
P=0.6582
3.0SL=0.7419
-3.0SL=0.5745
654321
1.0
0.9
0.8
0.7
0.6
0.5
Sample Number
Pro
port
ion
P Chart for In Phase
P=0.8145
3.0SL=0.8756
-3.0SL=0.7535
Strive to be
Level 5
64
654321
1.0
0.5
0.0
Sample Number
Pro
port
ion
P Chart for In Phase by Group
P=0.87503.0SL=0.9481
-3.0SL=0.8019
2
Control Limits for Requirements AnalysisPhase has no Overlap Whatsoever
Control Limits for Requirements AnalysisPhase has no Overlap Whatsoever
654321
1.0
0.9
0.8
0.7
0.6
0.5
0.4
Sample Number
Pro
port
ion
P Chart for In Phase by Group
P=0.6252
3.0SL=0.7106
-3.0SL=0.5398
2
Design Phase is More Interesting
Design Phase is More Interesting
654321
1.0
0.9
0.8
0.7
0.6
0.5
Sample Number
Pro
port
ion
P Chart for In Phase by Group
P=0.93733.0SL=0.9754
-3.0SL=0.8993
2
Control Limits for ImplementationPhase Have Significantly Changed
With Little Overlap
Control Limits for ImplementationPhase Have Significantly Changed
With Little Overlap
Range is much larger which widens the control limits.
However, looking at the scorecard, • only 1% of defects made it to test,
whereas before it was 14%. • Rework due to leakage was a
mere 13 days compared to 156 days!
Strive to be
Level 5
65
Make decisions based on data. Experience is only one input parameter and can steer you wrong
Let the Six Sigma tools work for you to narrow the scope and identify where to apply resources
A few simple process changes resulted in BIG impact to bottom line!
This project spun off others team which helped meet the customers’ needs
Quick look at charts show that the process is now in control
Greatest source of variation is now within the subgroups, not between them• Project to project variation is larger than phase to phase variation
The three phases that were concentrated upon improved greatly• Two of the three are world-class
Just measuring one aspect of the process does not tell the whole story• Finding a larger percentage of defects in phase may not be
cost effective
Conclusion
Strive to be
Level 5
66
New Challenge Was Presented
Strive to be
Level 5
67
478
239
14
280
0
100
200
300
400
500
600
Jul Aug Sep Oct Nov Dec
Requirements to review Requirements reviewed
Design reviewed Code reviewed
Item 17
Peer Review Tracking
Strive to be
Level 5
68
AS900 Peer Review Effort
0
10
20
30
40
50
60
70
80
90
100
1 5 9 13
17
21
25
29
33
37
41
45
49
53
57
61
65
69
73
77
81
85
89
93
97
101
105
Review Number
Effort per Review
Avereage Effort
Overall Peer Review Effort
12.6
Strive to be
Level 5
69
10050Subgroup 0
200
100
0Ind
ivid
ual V
alue
1
1 1 1 1 1
X=12.633.0SL=42.63
-3.0SL=-17.37
Plans# defects
Mod Hrs Prep Hrs
Peer Review Effort
Strive to be
Level 5
70
Peer Review Effort by Product Type
10050Subgroup 0
200
100
0
Ind
ivid
ual V
alue
1
1 111 1
1
X=8.7353.0SL=37.96
-3.0SL=-20.49
2 3 4 5 6Docs Reqs Design Code Test Change
Strive to be
Level 5
71
15105Subgroup 0
200
100
0
I nd
i vid
ua
l V
al u
e
1
X=24.52
3.0SL=93.70
-3.0S L=-44.66
Peer Review Effort by Product Type
2010Subgroup 0
40
30
20
10
0
-10
-20
I nd
i vid
ua
l V
al u
e
1
X=10.99
3.0SL=37.56
-3.0S L=-15.57
Docs Reqs
Design Code
Test Change
2010Subgroup 0
20
10
0
I nd
i vid
ua
l V
al u
e
X=6.904
3.0SL=16.81
-3.0S L=-2.997
252015105Subgroup 0
70
60
50
40
30
20
10
0
-10
-20
I nd
i vid
ua
l V
al u
e
1
11
X=13.16
3.0SL=43.83
-3.0S L=-17.51
32Subgroup 1
40
30
20
10
0
-10
-20
Ind
ivid
ua
l V
alu
e
X=11.69
3.0S L=35.41
-3.0S L=-12.02
252015105Subgroup 0
70
60
50
40
30
20
10
0
-10
-20
-30
I nd
i vid
ua
l V
al u
e
1
1
X=8.735
3.0SL=37.96
-3.0S L=-20.49
Strive to be
Level 5
72
10050Subgroup 0
300
200
100
0Ind
ivid
ual V
alue
1
1
X=18.323.0SL=60.85
-3.0SL=-24.20
Peer Review Number of Defects
Strive to be
Level 5
73
Peer Review Number of Defects by Product Type
10050Subgroup 0
300
200
100
0
Ind
ivid
ua
l Va
lue 1
1
X=15.103.0SL=43.65
-3.0SL=-13.45
2 3 4 5 6
Docs Reqs Design Code Test Change
Strive to be
Level 5
74
15105Subgroup 0
300
200
100
0
Ind
ivi d
ua
l V
alu
e
1
X= 27.00
3.0SL=107.1
-3.0SL= -53.14
2010Subgroup 0
60
50
40
30
20
10
0
-10
-20
Ind
i vi d
ua
l V
alu
e
X= 17.63
3.0SL=51.91
-3.0SL= -16.65
2010Subgroup 0
40
30
20
10
0
-10
Ind
ivi d
ua
l V
alu
e
X= 15.14
3.0SL=38.06
-3.0SL= -7.787
2010Subgroup 0
200
100
0
Ind
i vi d
ua
l V
alu
e
1
X= 23.59
3.0SL=83.87
-3.0SL= -36.69
32Subgroup 1
20
10
0
Ind
ivi d
ua
l V
alu
e
X= 8.000
3.0SL=19.97
-3.0S L= -3.968
2010Subgroup 0
70
60
50
40
30
20
10
0
-10
-20
Ind
i vi d
ua
l V
alu
e1
1
X= 12.22
3.0SL=39.30
-3.0SL= -14.86
Peer Review Number of Defects by Product TypeDocs Reqs
Design Code
Test Change
Strive to be
Level 5
75
198 Requirements* to peer review(predict 11 hrs and 18 defects/requirement review)
239 Design elements* to peer review(predict 7 hrs and 15 defects/design review)
464 Code elements* to peer review(predict 13 hrs and 24 defects/code review)
Reviews Remaining
* multiple elements may be reviewed in a single review
Strive to be
Level 5
76
478 481 492 495 502 502480 480 490
239 241
337
420
477502
1446
132
256
398
478
280
502 502
0
100
200
300
400
500
600
Jul Aug Sep Oct Nov Dec
Requirements to review Requirements reviewed
Design reviewed Code reviewed
Mission Accomplished!