translating the business needs into an improvement programme using cmm: a practical experience...
DESCRIPTION
This presentation has been given in 2003 at the Spider Conference to show how CMM could contribute to the business results of a large organization within PhilipsTRANSCRIPT
Translating the business needs into an improvement programme using CMM:
a practical experience within PHILIPS TV
Guy Van HooveldThanks to Hans Aerts, Baudewyn Meersseman and Geert Acke
for some of the presentation material
High End TV, Guy Van Hooveld 2
• History of SPI within PHILIPS
• High End TV – why SPI ?
• Achievements
• Recommendations
• New challenges
• Conclusions
High End TV, Guy Van Hooveld 3
Evolution of Embedded Software: 4 Software Stages
– Software Introduction stage
– Feature Expansion stage
– Mastering Complexity stage
– Complex Systems stage
Hardware-dominated industry
Software-dominated industry
High End TV, Guy Van Hooveld 4
Historic Overview• 1987 - 1991: Phase 0: Early Awareness Creation
Nov ‘87: Software Reference Model of the new “Center for Software Technology” mentions the “project” view
‘89: “Managing the Software Process” by Watts Humphrey ‘91 - ‘92: Growing SPI awareness within Philips
• 1992 - 1996: Phase 1: SPI Infrastructure Creation Jun ‘92: First SPI Guidelines published Jun ‘93: First SPI Steering Committee meeting Dec ‘96: Uniform approach within Philips for CMM certification
• 1996 - now: Phase 2: Monitor SPI Results 1997:
SPI Steering Committee with a clear assignment and budget CTOs clearly showed commitment to SPI
Dec ‘00: More than 50% of all software staff works in sites with CMM level 2 or higher
2003: number of CMM level 3 organizations is growing
High End TV, Guy Van Hooveld 5
• History of SPI within PHILIPS
• High End TV – why SPI ?
• Achievements
• Recommendations
• New challenges
• Conclusions
High End TV, Guy Van Hooveld 6
High End Market
High End products, High End expectations and targets
High End TV, Guy Van Hooveld 7
Business priorities
• Predictibility (shorter lead time, lower lead time and effort slip)
• Product quality (low field call rate)• Featuring (extended functionality)• Cost (reducing development cost, increasing
productivity)
• These priorities have to be translated into an improvement programme. Software is playing a key role.
High End TV, Guy Van Hooveld 8
Predictibility
• The world of consumer electronics is relying on introduction dates, selling seasons
• Being on time with the right product is paramount• Lead time, time slip are top priorities• Any delay is paid cash (market share, sales volume,
…)
High End TV, Guy Van Hooveld 9
Field call rate : what is acceptable?Field call rate : what is acceptable?
4 Cylinders +25.000 €
6 Cylinders + 40.000 €
8 Cylinders + 70.000 €
The higher the price, the higher the complexity, the higher the potential failure modes… ?
High End TV, Guy Van Hooveld 10
Expected quality in High End TV
• There is a correlation between product complexity (and size) and potential failure modes -> field call rate
• Does a BMW 7 user expect more problems than a BMW 3 user ?
• Does he/she accept more problems ?• Of course not :
QUALITY NEEDS ARE MUCH HIGHERin the HIGH END segment
High End TV, Guy Van Hooveld 11
Field call rate: what about this...?Field call rate: what about this...?
16Kb SW 200 €
64Kb SW/ 50 Hz.600 €
2Mb+ SW/100 Hz.1200 € to 8000 €
The higher the price, the higher the complexity, the higher the potential failure modes… ?
High End TV, Guy Van Hooveld 12
Evolution of Embedded Software: 4 Software Stages
– Software Introduction stage
– Feature Expansion stage
– Mastering Complexity stage
– Complex Systems stage
Hardware-dominated industry
Software-dominated industry
High End TV, Guy Van Hooveld 13
Software Performance Improvement
SW Introduction
Feature Expansion
Mastering Complexity
Open Systems
Process
Organization
Product
People
Technology
Architecture
High End TV, Guy Van Hooveld 14
Improvement Requirements per Stage
SW
Introduction Feature
Expansion Mastering
Complexity Open
Systems
Organization • Team: 1-3 • Single room
• Skill group • Team: 3-12
• Multi-site team, • Organization wide
• Organization wide, • Alliances
People • Electronic engineers
+ SW generalists + SW specialists + Architects
+ SW purchasers + SM managers
Technology • 4-8 bit µP • Assembler, C,
PL/M
• Req. Specs • 8-16 bit µP, • C
• Req./Design • 16-32 bit µP • C, C++, • RTK, components
• OS, • external applications
(DBMS, GUI, internet)
Architecture • None • Recycling of
product code • Proprietary
(closed) • Open Systems, • Standards
Process • Product
testing, • Basic CM
• CMM ≥ 2 • CMM ≥ 3 • Supplier Agreement
management
Evolution in Embedded SW Development
High End TV, Guy Van Hooveld 15
How much software does a shaver contain ?
Telephone exchange:1975: 256 Kbytes1990: 80 Mbytes
High-end TV:1985: 8 Kbytes2000: 2 Mbytes
Shaver:2000: 4 Kbytes2015: ?????
High End TV, Guy Van Hooveld 16
Software Growth
1
10
100
1000
10000
19
85
198
7
19
89
19
91
19
93
19
95
19
97
19
99
20
01
Year
So
ftw
are
Siz
e (
KB
yte
s)
TV VCR Double every 2 years
High End TV, Guy Van Hooveld 17
• History of SPI within PHILIPS
• High End TV – why SPI ?
• Achievements
• Recommendations
• New challenges
• Conclusions
High End TV, Guy Van Hooveld 18
Evolution time slip
CR / SR Timeslip
8%
46%
6%6%
0%0%
10%8%
42%
0%
10%
0%
6%
0%0%5%
0% 0%
0%
10%
20%
30%
40%
50%
Sep-98 Dec-98 Mar-99 Jun-99 Sep-99 Dec-99 Mar-00 Jun-00 Sep-00 Dec-00 Mar-01 Jun-01 Sep-01 Dec-01
Release date
% T
imes
lip
CR timeslip Target
01/2001
Decision making made possible: predictibility is also predicting delay.Peak in 2000 ? Product was not on the critical path and had the appropriate quality.The potential delay was predictable and has been taken into account.
High End TV, Guy Van Hooveld 19
Evolution post release defects density
Post-release defect density (faults/100 kbytes)
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
Jan-99 Apr-99 Jul-99 Oct-99 Jan-00 Apr-00 Jul-00 Oct-00 Jan-01 Apr-01 Jul-01 Oct-01 Jan-02
fau
lts/
kby
tes
High End TV, Guy Van Hooveld 20
Software quality (measured with FCR)
• 1998: top on the pareto list• 2000: major improvement visible• 2002: disappearing from the pareto• 2003: very few small incidents reappearing• 2004: ? (new platform)
High End TV, Guy Van Hooveld 21
Correlation pre-post release defectsPre Vs Post release defects
0
50
100
150
200
250
300
350
400
0 200 400 600 800 1000 1200
Pre-release defects
Po
st-r
elea
se d
efec
ts
We could expect a linear correlation, but reality shows that the number of post releasedefects is in fact more or less constant while complexity is growing.
High End TV, Guy Van Hooveld 22
Some achievements based on facts and figures
Results of SPI and of a platform approach:– Number of clusters (types of products) released has tripled
since 1998– Software product quality has significantly increased (field
calls reduced by a factor 10)
despite the following facts – Software size and complexity have quadrupled since 1998– No significant software headcount growth in High End TV
since 1999
High End TV, Guy Van Hooveld 23
• History of SPI within PHILIPS
• High End TV – why SPI ?
• Achievements
• Recommendations
• New challenges
• Conclusions
High End TV, Guy Van Hooveld 24
CMM priorities – step 1what to do first
• Basic processes• Seldom an issue (most organizations very quickly master those KPAs)• Configuration management• Requirements management• Basic planning and tracking (not necessarily CMM compliant)• Subcontract management (basic) if applicable
High End TV, Guy Van Hooveld 25
CMM priorities – step 2Increase control• Basic leverage
– Process quality• QA
– High level supportive quality officer– Escalation procedure– Issues tracked to closure– Subcontract management (extended) if applicable
• Estimates– Better planning and tracking (extended)
– Product quality (related to PE and PR KPAs)• Functional test team installed very early• Peer reviewing at all levels• Architecture• Guidelines (coding rules, design templates, …)
High End TV, Guy Van Hooveld 26
CMM priorities – step 3Mature organization
• Extended processes : • OPF-OPD – fully integrated strategy• ISM – PP/PT at a higher level: improve proactivity• IC – look outside software -> project management• TP – formalized
• Start thinking of CMM L4 and L5• Defects prevention first (focus on product quality)• Introduce statistical process control
High End TV, Guy Van Hooveld 27
CMM prioritiesWhat to do in parallel ?• To achieve the maturity levels, focus has to be put
on:– Improving the capabilities of the organization
• Install a SEPG group• Hiring higher level people• Organize training• Focus on interdepartmental awareness• Integrate in overall improvement plan• Adapt the organization structure
• To realize product strategy• Define and deploy future proof concept and architecture• Adapt the organizational model accordingly
High End TV, Guy Van Hooveld 28
• History of SPI within PHILIPS
• High End TV – why SPI ?
• Achievements
• Recommendations
• New challenges
• Conclusions
High End TV, Guy Van Hooveld 29
New challenges• Major new challenges for software:
– Cooperation between product development and IC related SW development (rely more on suppliers)
– “Digital” (huge increase of software size/complexity)• Leading to problems like:
– Customer-supplier relationship and software ownership– Customer-specific and common components in one architecture– Main software/DSP software integration
(technical integration of stacks and organizational cooperation between two existing large SW organizations with their own systems/structures/culture)
• Requiring solutions in:– Programme definition– Architecture– Process– Programme execution
High End TV, Guy Van Hooveld 30
New challenges (continued)
• Need to keep product quality under control• -> Process performance needs to be quantified and
analyzed: first candidates are Peer Reviews and Testing(improving the quality of defects filters at phase transitions)
• New kinds of testing are being deployed/extended• Home testing• Field testing• Stress testing• User perception
High End TV, Guy Van Hooveld 31
• History of SPI within PHILIPS
• High End TV – why SPI ?
• Achievements
• Recommendations
• New challenges
• Conclusions
High End TV, Guy Van Hooveld 32
Software Performance Improvement
SW Introduction
Feature Expansion
Mastering Complexity
Open Systems
Process
Organization
Product
People
Technology
Architecture
High End TV, Guy Van Hooveld 33
What is the key to implement change ?
• Stepwise approach to address business needs progressively
• Know what is wrong• Know what has to be changed first• Need to “Measure”, “Detect”• CMM L2, L3
• Relying on QA (quality officer)• Product quality information
• CMM L4, L5• Apply statistical process control
• Also focus on the other deployment axes
High End TV, Guy Van Hooveld 34
Different aspects are relatedHigh
Architecture and Technology level
Organization, Staff, and
Process level
problematic inefficient
high risk target
HighLow
High End TV, Guy Van Hooveld 35
Balanced Software Performance Improvements
High
Architecture and Technology level
Organization, Staff, and
Process level
problematic inefficient
high risk target
HighLow
SPISPI