Giant Plc 2009

1

Giant Plc.Barclays Plc.

Applying Six Sigma to Financial Software Development

Ali Raza KhanGiant Plc.

1 New Oxford Street, London ♦ United Kingdom(+44) 18443247700 ♦ [email protected]

© 2009 Giant Plc.

2


Motivation

Financial Software Financial Software DevelopmentDevelopment

• Large Opportunity for Improvement

• Approximately 25% of software projects are canceled

• Average project exceeds – Costs by 90% – Schedule by 120%

• Risk of project failure increases with size

© 2009 Giant Plc.

Six SigmaSix Sigma

• Well-defined improvement approach

• Impressive track record of achievements

• Adaptable

3


But Software Development is Not a Typical Application

DesignRequirements Testing IntegrationCoding Release

• Process Oriented, but– Inputs often ill-defined

– Outputs often difficult to fully evaluate

– Performance highly influenced by human factors (e.g., knowledge, skills, experience, etc.)

• Significant natural variation

© 2009 Giant Plc.

4


Key Factors in Software Project Failures

Risk Factor % of “MIS” Projects

Requirements Failures

Creeping Requirements 80%

Expectation Failures

Excessive Schedule Pressure 65%

Execution Failures

Low Quality 60%

Cost Overruns 55%

Inadequate Configuration Management

50%

5

Giant PlcBarclays Plc

Applying Six Sigma to Software Development

Design CodeIntegr-ation

Test

DMAIC

FuzzyFront-End

DFSS

BetaRelease

© 2009 Giant Plc.

6


Fuzzy Front End

Six Sigma DFSS

© 2009 Giant Plc

7


Balance the VOC and the VOB

Voice of the Customer

Voice of Business

8


VOC – Voice of the Customer

• Understand Internal and External Customers and Target Environment

© 2009 Giant Plc.

www.estm.biz 9

Building a Customer MatrixS

eg

men

ts

Types of Customers

Lead User

Demanding

Lost Lead

Had But Lost

U.S.

Europe

Asia

© 2009 Giant Plc.

10




• Identify, Characterize and Verify Critical to Quality (CTQ) Requirements– Interviews, focus groups, use cases, etc. – Preference surveys and Kano analysis

© 2009 Giant Plc.

11


Kano Analysis

• Dissatisfiers (or basic requirements)– “Must be” requirements– These features must be present to meet minimal

expectations of customers• Satisfiers (or variable requirements)

– The better or worse you perform on these requirements, the higher or lower will be your rating from customers

• Delighters (or latent requirements)– These are features, factors, or capabilities that go

beyond what customers expect, or that target needs customers can’t express themselves

© 2009 Giant Plc.

www.estm.biz 12

The Kano ModelH

ow

th

e C

ust

om

er F

eels

Neutral

Satisfier

Must-Be

Delighter

Level of Functionality Delivered

for a particular requirement

Low to None

High

Delighted

Very Dissatisfied

Satisfied, but I expect

it this way

I can live with it

© 2009 Giant Plc

13


VOC Output: Prioritized CTQsRequirement Use-case Kano Priority

Manage database interfaces Verifying data

content integrityS 4

Manage Network I/O

Moving client-server data

M 3

Optimizing data transfer

D 3.5

Provide real-time user access

Minimizing system response time

M 5

© 2009 Giant Plc.

14




• Identify, Characterize and Verify Critical to Quality (CTQ) Requirements– Interviews, focus groups, use cases, etc. – Preference surveys and Kano analysis

• Establish measures for CTQ requirements

© 2009 Giant Plc.

15


VOC Output: Fully Characterized CTQsRequirements Use-Case Kano Priority Measure

Manage Database Interfaces

Minimum Average Strong

Verifying data content integrity

S 4 ≤ 1 record/1,000 ≤ 1 record/10,000 ≤ 1 record/100,000

Manage Network I/O


M 3 100 records/min. 500 records/min. 800 records/min.


D 3.5 Hooks for user supplied

compression

Top 5 compression

schemes supplied

Top 10+ compression

schemes supplied and fully

integrated

© 2009 Giant Plc.

16


VOB - Voice of Business

• Analyze Design Options– Estimate customer satisfaction– Level of effort– Capability to deliver– Balance VOC and VOB

© 2009 Giant Plc.

17


Analyze Design Options

Requirement Use-Case Kano Priority Base Full Base

Effort

Full Effort

Manage database interfaces


S 4 1 3 1000 1500

Manage Network I/O


M 3 1 3 5500 7500


D 3.5 1 3 12000 18000

Customer Sat. Score

Effort Score

Design Options Level of Effort

© 2009 Giant Plc.

18



Requirement Use-Case Kano Priority Base Full BaseEffort

Full Effort



S 4 1 3 1000 1500

Manage Network I/O


M 3 1 3 5500 7500


D 3.5 1 3 12000 18000

Customer Sat. Score

Effort Score

= F(Kano, Priority, Feature Level)

© 2009 Giant Plc.

19



Requirement Use-Case Kano Priority Base Full BaseEffort

Full Effort



S 4 1 3 1000 1500

Manage Network I/O


M 3 1 3 5500 7500


D 3.5 1 3 12000 18000

Customer Sat. Score

Effort Score

= ∑ Effort Estimates

© 2009 Giant Plc.

20


Concept Selection

4 3

1 2

CustomerSatisfaction

Eff

ort

© 2009 Giant Plc.

21


Computing Productivity

Historically, for each project we should know Size, Effort, and Duration

)(

)(/

/3431

*

(SLOC)

yearsDurationB

StaffYearsEffort

SizePP

© 2009 Giant Plc.

22


Schedule Compression

3)()(

torelates

yearsDurationStaffYearsEffort

MBI

MBI Buildup Rate Equation Output

1 Slow 7.3

2 Mod. Slow 14.7

3 Moderate 26.9

4 Rapid 55

5 Very Rapid 89

Manpower Buildup Index, MBI

© 2009 Giant Plc.

www.estm.biz 23

Rayleigh Curve

Rayleigh Summary

0.0

2.0

4.0

6.0

8.0

10.0

12.01 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Time Interval

Eff

ort

0.0

20.0

40.0

60.0

80.0

100.0

120.0

140.0

160.0

Def

ects Effort

Defects

© 2009 Giant Plc.

24


MBI = 1 (Slow) Concept 1

Duration (months) 15.2

Effort (staff months) 77.3

Released Defects 14.1

Effort Cost $966,250

Duration Adjustment

Defect Repair Cost $239,700

Net Value $19,482,050

MBI = 3 (Moderate) Concept 1

Duration (months) 13


Released Defects 267

Effort Cost $1,492,500

Duration Adjustment $1,400,000


Net Value $20,141,600

Balancing VOC and VOB

Business Value $20,000,000

Feature Value $688,000

© 2009 Giant Plc.

25


Balancing VOC and VOB

MBI = 5 (Very Rapid) Concept 1

Duration (months) 11.7






Net Value $18,665,650

MBI = 3 (Moderate) Concept 1

Duration (months) 13






Net Value $20,141,600

Business Value $20,000,000

Feature Value $688,000

© 2009 Giant Plc.

26


VOB - Voice of Business

• Analyze Design Options– Estimate customer satisfaction– Level of effort– Capability to deliver– Balance VOC and VOB

• Select Concept and Approach– Flesh out concept

• QFD• FEMA

– Verify and refine approach• Defect analysis• Schedule simulation

© 2009 Giant Plc.

www.estm.biz 27

Capability to Deliver on TimeProbabilistic Scheduling

Frequency Chart

Certainty is 94.90% from -Infinity to 289.17 days

.000

.007

.014

.021

.028

0

7

14

21

28

250.00 262.50 275.00 287.50 300.00

1,000 Trials 4 Outliers

Forecast: F52

How much confidence should we have in the schedule?… At a 95% confidence level

• latest mid March, 2003 (+ 43 days)• earliest mid January, 2003 (- 15 days)

Upper Spec Limit

(USL)

© 2009 Giant Plc.

28


Process Improvement

MeasureDefine Analyze Improve Control

Standard Six Sigma DMAIC Process

© 2009 Giant Plc.

29


Application to Software

DesignRequirements Testing IntegrationCoding Release

Requirements ReleaseDevelopmentX

Prerequisites:Processes must be well defined

© 2009 Giant Plc.

30


DMAIC Example

• Problem Statement– Post release maintenance has increased by 30%

since the end of last fiscal year and is now limiting new product development.

• Goal Statement– Reduce post release maintenance by 40% by the

end of Q4’2003.

© 2009 Giant Plc.

31


Measure – Data Collection

• Total Problems Fixed Prior to Release Per Project– Pre-Release Defects: defects found and fixed during

development and testing

• Total Post Release Problems Per Project– Released Defects: defects reported by customers

• Types of Post Release Problems– All projects

– Per project

© 2009 Giant Plc

32


Analysis

To

tal P

re-R

ele

ase

De

fec

ts

Project Size (LOC)

x x

x xx

xx

x

x

xx

x

Pre-Release Defects = f(Size)

© 2009 Giant Plc.

33


Analysis

Esc

aped

Def

ects

Pre-Release Defects

x

x

x

x

x

xx

xx

x

x

x

• Escaped defects proportional to pre-release defects – No significant variation in Defect Containment Effectiveness

• DCE = Pre-Release Defects/(Pre-Release Defects + Escaped Defects)

© 2009 Giant Plc.

34


Analysis

40%29% 26%

5%

Code

DesignReq’ts

Test

• Most Escaped Defects are Code Related

© 2009 Giant Plc.

35


Improve

• Improve the Effectiveness of Code Inspections– Factors

• Size of unit (LOC)

• Preparation time (LOC/hour)

• Inspection time (LOC/hour)

• Number of reviewers

– Measure• Number of identified defects

© 2009 Giant Plc.

36


Improve

• Improve the Effectiveness of Code Inspections– Conduct DOE

• Determine most effective combination of factors

– Verify DOE results• Pilot test using real project

© 2009 Giant Plc.

37


Control

• Establish Performance Standard for Code Inspections– Defects/KLOC

• Monitor Performance– Take action when unacceptable performance

observed

© 2009 Giant Plc.

38


Questions

© 2009 Giant Plc.

Date post:	02-Nov-2014
Category:	Documents
Upload:	alirazakhan83
View:	1,058 times
Download:	0 times

Giant Plc 2009

Documents