Performance Benchmarking Consortium
12/12/2006
SOFTWARE PRODUCTIVITY RESEARCH
Minds Meeting Metrics The SEI Performance Benchmarking Consortium December 12, 2006 – Boston SPIN Michael A. Bragen, Managing Partner Chair, PBC Specification Committee Chair Software Productivity Research, LLC http://www.SPR.com
SPR • P: +1.781.273.0140 F +1.781.273.5176 • www.spr.com
Presentation Topics
Introduction
Overview – The Performance Benchmarking Consortium Motivation & Benefits Ground covered
PBC Working Teams
Concept of operations Voice of customer Specification
Next Steps
Near-term goals How you can participate
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
1
Performance Benchmarking Consortium
12/12/2006
Motivation
Organizations want a way to gauge their performance and to compare their performance with others in their industry.
Data on project performance is needed to demonstrate the impact of process improvement.
Benchmarks
provide a reference point for interpreting performance facilitate interpretation by setting specifications Cost for how performance measurements are collected Schedule
Quality
Customer satisfaction
Draft version 0.1
Is There Community Interest?
During June, 2006, the PBC conducted an initial survey to assess the voice of the customer.
14 can authorize sharing of organization’s performance data
63 part of a team who can authorize sharing of organization’s performance data
Results of 2006 PBC Survey assigns a high degree of value to software • Sample size = 800 project performance benchmarks • Response Outcome ≈ 25%
12 out of 14 or assigns a high degree of value to software project performance benchmarks 85.7% 57 out of 63 or 90.5%
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
2
Performance Benchmarking Consortium
12/12/2006
What is Benchmarking?
Term
Description
Benchmark
To take a measurement against a reference point.
Benchmarking
A process of comparing and measuring an organization with business leaders anywhere in the world to gain information which will help the organization take action to improve its performance.
The Benchmarking Management Guide American Productivity and Quality Center
Draft version 0.1
Types of Process Benchmarking
Term
Description
Internal studies
Compare similar operations within different units of an organization.
Competitive studies
Target specific products, processes, or methods used by an organization’s direct competitors.
Functional or industry studies
Compare similar functions within the same broad industry or compare organizational performance with that of industry leaders.
Generic benchmarking
Compares work practices or processes that are independent of industry.
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
3
Performance Benchmarking Consortium
12/12/2006
Software Benchmarking Model
Historical Base of Project Experience Draft version 0.1
The Three-Legged Stool
•
Holistic view of software project performance factors - interacting effects
Project Duration
Project Quality
Project Effort (cost)
•
Key premises: 1) Improvements to one leg always affect performance of others 2) Higher organizational maturity enables better overall performance 3) Priority setting is a business, not technical decision
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
4
Performance Benchmarking Consortium
12/12/2006
Benchmark Derivation and Modeling
MEASURING
QUANTITATIVE DATA
QUALITATIVE DATA
Size Effort Schedule Documentation Defects
Personnel Process Technology Environment Products
Productivity Rates Quality Levels
ASSESSING
IMPROVING
Current state
Project Profiles
Best Case Models
Gap analysis
Influences
Draft version 0.1
SPR Knowledge Base 11,307 projects (2006)
Project Size
Project Classification Large 23% Enhancement 55%
Project Type Systems 35%
Medium 50%
MIS 42% Small 27% New 30% Maintenance 15%
Miscellaneous 16% Commercial 7%
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
5
Performance Benchmarking Consortium
12/12/2006
Linear regression analysis
• SPR uses r2 value statistical analysis to determine best-fit curves around the peer projects • The comparative quantitative data charts use these curves to present comparisons between each project (data point) and statistically significant subsets of the SPR Knowledge Base
Draft version 0.1
Benchmark Comparison Terminology – Peer Worst and Best -
•
The SPR analysis is based upon the following benchmark comparison values:
Peer Best: largest (maximum*) value in the range Peer Worst: smallest (minimum*) value in the range
Actual project data points from peer set
1,000 Maximum value
100
μ = ∑(1..n) / n
10
Minimum value
1 1
4
7
10 13 16 19 22 25 28 31 34 37 40 43 46 49
* This represents the opposite for measures with negative correlation to larger magnitude. Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
6
Performance Benchmarking Consortium
12/12/2006
Project Productivity – Variance to Peer Mean Values
Project Productivity Comparison All Platforms Type 1
Type 2
Type 3
Type 4
Project Quality Comparison All Platforms
Type 5
Type 1
-
Type 2
Type 3
Type 4
Type 5
-
Code Defects per 1000 FP
20
100
Obs
150
BIC Peer Best Peer Worst
200
40 60 80 100
Obs BIC Peer MEAN Peer Worst
120 140 160
Draft version 0.1
Effort Comparisons APPLIED TIME (Effort Months) UTE/Endesa Projects Unregulated Utility Regulated - Web
2 Tier C/S
Project durations areRegulated consistently Utility lower than expected ranges 10240 5120 2560 1280 Function Poiints
Hours per FP
50
640
R
320 O
160
4
80 Y D
40
X
P
G 3
20 10
M F
W
N U I V L
BH
1 K
A
C2S
4
16
64
256
1024
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
7
Performance Benchmarking Consortium
12/12/2006
Productivity Rate Comparisons PRODUCTVITY (Function Points / Effort Month) UTE/Endesa Projects
2 Tier C/S
Observed values Unregulated Utility from sample Regulatedset Utilityare Regulate - Web from Year 2004 and 2005 only. 10240 5120 2560
Function Points
1280 640
R
320 O 4
160
M
F
80
W D
X
Y
G
40 A
P 1
K
20
S
3 HB 2C
N
U
V
I
L
10 0
2
4
6
8
10
12
14
Function Points per Effort Month
Draft version 0.1
Defect Removal Efficiency Defect Removal Efficiency 2 Tier C/S Regulated Utility
Unregulated Utility Regulated - Web
10240 5120 2560
Function Points
1280 640 320 160 80 40 20 10 70.00
80.00
90.00
100.00
Defect Removal %
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
8
Performance Benchmarking Consortium
12/12/2006
Composite Quantitative Overview Composite analysis of key performance UTE Project Comparisons factors shows strengths offsetting A B C D K N S V Y 2 H I L weaknesses in theU environment M
R
W
1
3
F
G
O
P
X
4
500% 2
400% Better than similar projects
300%
V
U
G
200%
P
Y
100% U
D C
R W
2
O P
0% Worse than similar projects
S VY A
-100%
I 4
M
3 1 F
K
D N
R L
VY
X BC S K A
2
H
M
W 1 F G
X
I H L N U VY
B
P O
3
C
G
HI I
VY
L
R
P
1
H
K S W
G F O X
B C
2
M
I
4 F
L
M
W 1
D
U
13 W R
O X
X F
R L
4
A
K
4
-300%
G
H
-400% -500%
P O
3
S
2
N
-200%
D N
M
D 4
U 3
A
B
Time to Market
Productivity
Staffing (FTE)
Effort
Documentation
Draft version 0.1
Risk / Value analysis Risk / Value Placement
LOW
Risk
HIGH
LOW
Value
HIGH
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
9
Performance Benchmarking Consortium
12/12/2006
Qualitative Analysis – Environmental Attributes Detail
Draft version 0.1
Attribute Rating Scale
RATING
RANGE
Leading Edge
1.00 - 1.50
Above Average (Competitive)
1.51 - 2.50
Average
2.51 - 3.50
Below Average
3.51 - 4.50
Deficient
4.51 - 5.00
Note: All Attribute data collected for this analysis represents the responses of the project teams based upon their perception of each factor at the beginning of the project.
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
10
Performance Benchmarking Consortium
12/12/2006
Attributes Analysis by Level Environmental Attributes
Personnel
Technology
Process
Environment
Product
Management
Developers
Application experience
Analysis experience
Selected views
Design experience
Quality
Users
Draft version 0.1
Development Process Factors – (Process)
Development Process
Leading Edge
Improvement here should aid the clarity and stability of requirements. Competitive
Average
Below Average
Deficient
Difference is related to interpretation of prototyping.
Improvement in requirement process should balance and improve this factor. Requirements Clarity and Stability
Requirements Methods
Prototyping Methods
Analysis Methods
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
11
Performance Benchmarking Consortium
12/12/2006
Improvement Priorities
•
This analysis provides a view of self-identified gaps and objectives in the development teams that pertain to organizational capability and maturity attributes SPR considers important for performance (quality and productivity levels) Scores are based upon interviews and qualitative data collection activity conducted with development teams and managers.
Draft version 0.1
The PBC is Born
Objectives •
During April 2006, SEI launched a vendor and industry collaboration on benchmarking software project performance.
• Provide tools and credible data for goal-setting and performance improvement • To combine benchmark data from multiple repository sources thereby creating a superset of information for benchmark and/or performance comparison
Value • Establish specifications for the collection and comparison of benchmark data from different vendor sources • Allow companies to leverage existing data to help them establish and achieve their business goals
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
12
Performance Benchmarking Consortium
12/12/2006
Current PBC Members as of October 2006
•
Consortium members are leaders in software measurement and benchmarking from consultancies, industry, and academia.
Galorath
QSM
ISBSG
Raytheon
Lockheed Martin
SEI
Motorola
SPR
Oracle
STTF
PRICE Systems
The David Consulting Group
PRTM
University of Ottawa
* International Software Benchmarking Standards Group (ISBSG) ** Software Technology Transfer Finland (STTF)
Draft version 0.1
Broader Data Provided by PBC
Vendor Vendor Vendor Vendor Vendor 1 2 3 4 5
PBC
+ + + + + + +
Software size Defect density Defects by phase Productivity Schedule predictability Effort Customer satisfaction
Granularity of Data M O R E
LESS
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
13
Performance Benchmarking Consortium
12/12/2006
Broader Data Set Provided by PBC
PBC Member #1 Member #2 Member #3 Member #4 Member #5 Member #6 Member #7 Member #n
Draft version 0.1
PBC Provides Broader Coverage of Application Domains
Education Insurance Health and Medicine Military Systems Financial & Banking Manufacturing Electronics Communications
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
14
Performance Benchmarking Consortium
12/12/2006
Presentation Topics
Introduction
Overview – The Process Benchmarking Consortium Motivation & Benefits Ground covered
PBC Working Teams
Concept of operations Voice of customer Specification
Next Steps
Near-term goals How you can participate
Draft version 0.1
Ground Covered - 1
Kick-Off Workshop at SEI [April 19-20, 2006]
14 presentations by workshop attendees; Discussion of current benchmarking issues and ways to address them.
Brainstorming issues on how to proceed.
Initiative to conduct survey to obtain community input on factors most likely to affect software project performance.
Performance Benchmarks Consortium (PBC) is born. What is performance measurement? What makes a benchmark good and useful? What constitutes valid data if you are interested in learning about your range of results in comparison to other benchmarking companies? How should performance measurements be categorized?
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
15
Performance Benchmarking Consortium
12/12/2006
Ground Covered - 2
PBC Workshop [June 28-29, 2006]
Initial survey addressing performance factors results presented
Planning and goal-setting
Soliciting member input on the PBC business concept
Initial concepts about PBC products and services
Setting up work teams
Draft version 0.1
Ground Covered - 3
PBC Workshop [October 4-5, 2006]
Work team status reports and planning
Discussion of ConOps (Concept of Operations)
Presentations by Member Companies
Discussion of PBC Measurement Specification
Selection of initial set of performance influence factors and performance indicators that will populate version 1 of the PBC repository.
PBC 2006 planning and 2007 goal-setting
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
16
Performance Benchmarking Consortium
12/12/2006
PBC Member Alignment is Happening
Work during 2006 as focused on • Developing a common understanding of the terrain
2007
• Developing common goals • Cultivating a shared commitment
October, 2006
April, 2006
Draft version 0.1
Presentation Topics
Introduction
Overview – The Process Benchmarking Consortium Motivation & Benefits Ground covered
PBC Working Teams
Concept of Operations Voice of Customer Specification
Next Steps
Near-Term Goals How you can participate
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
17
Performance Benchmarking Consortium
12/12/2006
Current PBC Teams
ConOps Team
• Business planning • Communication planning • Concept of operations
Voice of Customer Team
• Soliciting community input • What do people need?
Specification Team
• Performance measurement definitions & guidance
Pilot Implementation Team
• Testing the solution components
Draft version 0.1
Concept of Operations - 1
Subscriber Performance Data
Member Assets PBC Member Companies
SEI
Subscriber Organizations
PBC Repository
Members provide services & reports to their internal or external customers using PBC Assets in addition to, or in combination with, their own
Customized reports
Measurement Specification
PBC Performance Reports
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
18
Performance Benchmarking Consortium
12/12/2006
Concept of Operations - 2
3
PBC Member Assets
1
Measurement Specification
PBC Measurement Specification
4
Subscriber Organizations
2 PBC Repository
• Using specification allows subscribers to make valid comparisons. • Becomes an ISO standard?
Subscriber Inputs 100
100 Total Staff Hours Expended (X1000)
94
90
90
80
80
Actual 70
70 60
60 Frequency 50
50
Planned
40
40 30
30
20 20
14
20 10
10
10
8 4
0
0 Jan
Feb
March
April
May
Month
June
July
August
Sept
No Response Don't Know or N/A
Strongly Disagree
Disagree
Agree
Strongly Agree
Response
Draft version 0.1
Concept of Operations - 3
The SEI: • authors Meas. Spec. • houses the repository • provides website & admin • authors PBC Reports • provides communication support PBC Member Companies
Subscriber Organizations SEI
PBC Repository Subscribers submit performance data that adheres to Meas. Spec.
Member companies: • provide assets to stock the repository • collaborate on Meas. Spec. • pay fee to sustain operations • provide training & consultation
Measurement Specification
They pay fee for: • PBC Performance Reports • Customized reports
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
19
Performance Benchmarking Consortium
12/12/2006
Concept of Operations - 3 Benchmarking From PBC Performance Report
PBC
PBC Repository
Member #1 Member #2 Member #3
Client 1
Member #4
Client 2
Member #5
Our client provided that measure. We can show you how they did it. We are
Client 3
here.
But we want to be here.
Member #6 Client 4 Member #7
PBC Subscriber
Client 5 Member #n
Draft version 0.1
Presentation Topics
Introduction
Overview – The Process Benchmarking Consortium Motivation & Benefits Ground covered
PBC Working Teams
Concept of Operations Voice of Customer Specification
Next Steps
Near-Term Goals How you can participate
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
20
Performance Benchmarking Consortium
12/12/2006
Voice of Customer Team
This team if focused on ensuring that the voice of the customer is represented as PBC products are planned for development. • Conducted a survey to obtain community input about factors that influence software project performance • Planning a birds-of-feather session to solicit community input to influence the development of PBC products and services
Draft version 0.1
Survey:
What Are the Key Factors Affecting SW Project Performance?
• Management and leadership
• Project types development
• Project organization environment
• Application domain
• Analysts' functional knowledge
• Technical complexity
• Developer skill level
• Use of development methodology
• Process maturity
• Product architecture
• Team dynamics
• Project risks
• Team size
• Project technology – language & tool effectiveness
• Volume of staff turnover • External customer relationship • External integration complexity • Business domain
• Newness of development platform • Platform volatility • Project technology – familiarity with • Size
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
21
Performance Benchmarking Consortium
12/12/2006
Sample Characteristics & Response Rates
Population Size
Sample Size
SEI Cust. Relations
6,398
500
DCG
2,016 8,414
Subpopulation
Total
Actual Sample Size
Response Rates RR1
RR2
407
20.9%
24.0%
500
412
22.8%
27.7%
1,000
819
21.9%
25.9%
Due to email bounce-backs and ineligible respondents. RR1: RR2:
Minimum Response Outcome. Excludes those who did not complete the entire questionnaire. Maximum Response Outcome. Includes those who partially completed the questionnaire.
Draft version 0.1
Top-10 Factors
“Very Large” Impact
“Very Large” or “Large” Impact
1 Management and leadership 45.3%
1 Management and leadership 87.6%
2 Developer skill level 42.9%
4 Project organization environment 83.6%
3 External customer relationship 31.7%
9 Team dynamics 79.9%
4 Project organization environment 31.3%
2 Developer skill level 77.5%
5 Product architecture 28.6%
3 External customer relationship 74.7%
6 Project risks 28.0%
5 Product architecture 74.7%
7 Process maturity 27.9%
7 Process maturity 74.2%
8 Technical complexity 26.9%
6 Project risks 73.6%
9 Team dynamics 24.9%
10 Use of development methodology 72.5%
10 Use of development methodology 23.6%
8 Technical complexity 65.9%
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
22
Performance Benchmarking Consortium
12/12/2006
Presentation Topics
Introduction
Overview – The Process Benchmarking Consortium Motivation & Benefits Ground covered
PBC Working Teams
Concept of Operations Voice of Customer Specification
Next Steps
Near-Term Goals How you can participate
Draft version 0.1
Specification Team Work
The Spec Team is working to develop common definitions.
Performance Influence Factors
Performance Indicators
• Process maturity
• Defect density
• Application domain
• Time-to-market
• Stability of requirements
• Schedule predictability
• Size
• Productivity index
• Project type
• Project delivery rate
• Team size • Developers' functional knowledge
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
23
Performance Benchmarking Consortium
12/12/2006
Specification Team – Building Common Definitions
Example: Time-to-market measure
S/W Project A Meas. Definition A S/W Project B Meas. Definition B
How can you compare?
S/W Project C Meas. Definition C S/W Project D Meas. Definition D
The Specification
A cool, refreshing common definition for “time-to-market” Team that permits valid comparisons between software projects
Draft version 0.1
PBC Measurement Specification
PBC Member Assets Measurement Specification
PBC Measurement Specification
Subscriber Organizations
4
2 PBC Repository
• Using specification allows subscribers to make valid comparisons. • Becomes an ISO standard?
Subscriber Inputs 100
100
94
90
90
Total Staff Hours Expended (X1000)
3
80
80
Actual 70
70 60
60 Frequency 50 50
Planned
40
40 30
30
20 20
14
20 10
10
10
8 4
0
0 Jan
Feb
March
April
May
Month
June
July
August
Sept
No Response Don't Know or N/A
Strongly Disagree
Disagree
Agree
Strongly Agree
Response
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
24
Performance Benchmarking Consortium
12/12/2006
Presentation Topics
Introduction
Overview – The Process Benchmarking Consortium Motivation & Benefits Ground covered
PBC Working Teams
Concept of Operations Voice of Customer Specification
Next Steps
Near-Term Goals How you can participate
Draft version 0.1
Near-Term Goals
The PBC will
create a set of process specifications for the consistent and meaningful collection, analysis, and dissemination of comparative performance benchmarks for software projects
develop a data model that will facilitate the aggregation and comparison of data from different sources
pilot test solution components
develop version 1.0 of the PBC data repository
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
25
Performance Benchmarking Consortium
12/12/2006
Are You Interested?
The PBC will be expanding membership during 2007. Individuals who become members during 2007 will participate in proof-of-concept trials and provide feedback on the PBC Concept of Operations. If your organization is interested in sponsoring you as a member or if you are interested in becoming a subscriber of future PBC products and services, then please send email to Mark at
[email protected] or Michael at
[email protected] Individuals who contact us will be added to our database. We will periodically email about progress and future plans for the PBC. Thank you
Draft version 0.1
Acknowledgements
Kate Armel
QSM
Michael Bragen
Software Productivity Research
Khaled El Emam
University of Ottawa
Eric Finch
PRTM
SEI Anita Carleton
Robert Floyd
Raytheon
Robert Ferguson
Pekka Forselius
STTF
Diane Gibson
David Garmus
DCG
Dennis Goldenson
Peter Hill
ISBSG
Mark Kasunic
Tim Hohmann
Galorath
Oksana Schubert
Thomas Lienhard
Raytheon
Robert Stoddard
Larry McCarthy
Motorola
Dave Zubrow
Arlene Minkiewicz
PRICE Systems
Kristal Ray
Oracle
Suresh Subramanian
PRTM
Bob Weiser
Lockheed Martin
Draft version 0.1
© 2006 Carnegie Mellon University © 2006 Software Productivity Research, LLC
26