You are on page 1of 32

SOFTWARE PRODUCTIVITY CONSORTIUM

Integrating Six Sigma


and the CMMI

David N. Card

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Topics

• Background
• Six Sigma and the CMMI
• Design for Six Sigma and Software
Development
• Putting It All Together
• Challenges

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Background

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Problem Statement

• Adoption of Six Sigma is increasing among


Member Companies who already employ
CMM/CMMI-based software process
improvement.
• They are superficially different approaches:
– Different language and terminology
– Different consultants
– Different professional societies
• Does Six Sigma obviate CMM-based process
improvement?
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Some Historical Influences

Shewhart
SPC
Six Sigma

Fisher
DOE
Crosby
COQA CMMI

Paulk
CMM

Humphrey
Box
85Qs
EVOP

1900 1920 1940 1960 1980 2000

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Six Sigma and CMMI

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Quality Management Maturity Grid

CMM/CMMI Translate the QMMG into


the Software and Systems Domains

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
What’s Six Sigma?
Spec. Limits

• Another quality slogan, for many


σ

• A business-focused philosophy
based on statistical thinking
Performance Measure
• A goal for process capability
– Very small probability of unacceptable result (approximately 3
defects in a million to 2 defects in a billion)
– Only arrived at by understanding and managing process
performance
– Usually only meaningful for a process element or activity
• An integrated set of established techniques including
statistical process control, design of experiments, quality
function deployment, benchmarking, etc.
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
σ on Benchmarking

• Essential to becoming “best in class”


• Typically need to benchmark with many
different organizations to cover all processes
• Can be done via literature searches
• Should be a continuing activity
• CMM provides synthetic benchmark of best
practices for software engineering

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Control Charts Provide a Useful
View of Many Processes

Inspections of State Transition Diagrams

Upper Limit

Center
Actuals

From D.Card, Controlling the Object-Oriented Design Process, CNRC Conference on


SOFTWARE
Quality Assurance of Object-Oriented Software, February 2000 PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Significant Differences
Six Sigma CMM/CMMI
Assumes processes have been Focus on defining management and
identified and defined technical processes early
Doesn’t distinguish organizational Organizational process definition
standard and project processes used to capture best practices
Emphasis on training to motivate Emphasis on infrastructure to
and communicate skills ensure key processes addressed
Reliance on statistical methods to Statistical approach intended often
manage performance not implemented
Focus on learning from internal Additional mechanisms to leverage
experience and data external technology
Prioritization of efforts based on Link to strategic planning weak and
business payoff often ignored
Certification of individual Certification of assessors and
practitioners, not organizations organizations, not practitioners
David N. Card, Sorting Out Six Sigma and the CMM, IEEE Software, July 2000 SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Practical Implications
• Adopting a Statistical Process Control-based
approach to Level 4 facilitates the transition to Six
Sigma
• Six Sigma is difficult for Level 1 organizations to
implement
• Six Sigma provides an alternative (not a short cut) to
pursuing Level 5 for Level 3 organizations seeking to
instill continuous improvement
• Incorporating Six Sigma techniques helps
organizations working towards Level 4 and 5 to
deliver the best business results
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Design for Six Sigma and Software
Development

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Design for Six Sigma

• Moving past five sigma often requires


redesigning processes and products
• Two strategies commonly applied:
– Use standardized and proven parts (i.e., reuse
and COTS)
– Minimize complexity in process (e.g., fewer steps)
and product (i.e., increase producibility)
• Implies concurrent design of process and
product (e.g., tailoring of organizational
process)
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Traditional Manufacturing Context

Design Design Execute


Product Process Process

Product Process Deliverable


Design Design Products

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Typical DFSS Context
DFSS
(Process)

Design Six Sigma


Process

Process
Design

Deliverable
Execute
Design Products
Process
Product

Product
Design

DFSS
(Product)
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Two Related Objectives

• Design the process from the start to achieve


Six Sigma
• Design the product to facilitate the process –
minimize breakage during production
 DFSS is not a software development
methodology, but its concepts and techniques
can be incorporated into one!

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Producibility in Software

9
10

8 B A
G
7
H F
6

5
C
Defects 4
E
Per D
KSLOC 3 r = 0.83
p(r = 0) < 0.02
2

25 30 35
Design Complexity
SOFTWARE
From D.N. Card, Measuring Software Design Quality, Prentice Hall, 1990 PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Putting It All Together

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
There’s No Easy Answer!

• Which software development approach


should I adopt?
– CMM/CMMI
– Extreme Programming
– Iterative Development
– DFSS
– Etc.
• Processes (methods, tools, people) provide
competitive advantage – have to be adapted
to the organization
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Key Concepts for Integration

• Each paradigm addresses a different issue


• CMM/CMMI describe what must be done, not
how to do it
• Six Sigma and DFSS provide generally
applicable techniques
• Software and systems development
methodologies define the detailed “problem
appropriate” steps to accomplishing work

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
First Step Towards Integration
Six Sigma DFSS
(Process) Support

Process Engineering
Audits, etc.
Management Extreme
Programming, etc.
DFSS
(Product)

Software
System Product
Design
System
Product
Project Earned Manufacturing
Management Value, etc. (May be just
duplication)
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Six Sigma in Support Processes
Regression
These are not all the possibilities… Analysis is a
EMEA is a mechanism for
mechanis CAR
m for CAR

Control Chart
Pugh Matrix is is a
a mechanism mechanism for
for DA&R M&A.
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Six Sigma in Project Management

These are not all the possibilities…


Control Chart
is a mechanism
for OPM.

Control Chart
is a mechanism
PM&C.
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Six Sigma in Process Management
Deployment Flow
Chart is a These are not all the possibilities…
mechanism for OPF
and OPD.

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Six Sigma in Engineering
These are not all the possibilities…

Quality Function
Deployment can be
a mechanism for
RM, RD, and TS.

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Challenges

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Common Problems

• Adoption of Statistical Thinking


• Statistical Issues
• Conceptual Models
• Focus on the score
– CMM Level
– “Sigma” rating
• Reluctance to recognize (and measure) the
magnitude of software rework (Typically 30-
50%)
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Adoption of Statistical Thinking
• Ambiguous use of “obvious” measurement
terms
• Unfamiliarity of statistical terms, concepts,
and techniques
• Lack of good engineering process data
– Length of time it takes to produce data
– Difficulty in performing and scaling experiments
• “Desire for Perfection”
– Convenient excuse for resisters
– Overload produced by enthusiasts
• Lack of software-oriented statistics training
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Conceptual Models
• Design of Experiments
– C. 1900 (agriculture)
– Multiple trials following experimental design
– Extraneous factors eliminated
• Evolutionary Operation
– C. 1950 (chemical processes)
– Systematic variation of key parameters
– Normal operational conditions
• Statistical process control (c. 1920,
manufacturing) accompanies both
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Statistical Issues

• Distributional and data concerns


– Control Charts:
XR chart (typical for manufacturing)
XmR and U charts
– ANOVA and Regression
Parametric (based on normal distribution)
Nonparametric and distribution-free analyses
• Elaborated versus assembled product

SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM
Summary

• Six Sigma and CMM/CMMI-based process


improvement are complementary:
– Six Sigma strengthens analysis capabilities
– CMM provides organizational structure
• Six Sigma provides a new package for existing
quality and statistical techniques
• Successful integration of these techniques requires
expertise beyond just “assessing”
• Limited research in addressing the issues that limit
the application of Six Sigma to software and systems
SOFTWARE
PRODUCTIVITY
Copyright © 2001, Software Productivity Consortium NFP, Inc. All rights reserved. CONSORTIUM

You might also like