UTSQ Presetation (slides)
| UTSQ Presentation | Download: |
|---|---|
| Pages: 1 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 | |
| Page: 1 | Top Next |
![]() Unifying Theory of Software Quality (UTSQ)
Because all Software is a series of Economic Events and Responses
| |
| Page: 2 | Prev Top Next |
![]() Software is an instantiation of a concept believed to deliver economic value
Requirements are a representation Code is a representation Test suites are a representation Documentation is a representation
Concept = Hypothesis
| |
| Page: 3 | Prev Top Next |
![]() Motivation
Why A New Theory
| |
| Page: 4 | Prev Top Next |
![]() Reality Hits in Quality Processes
Many direct hits between what is taught and what
is done
ISO certification is required in Telco equipment Software Engineering Institute (SEI) is a driving force in software development processes for DoD. Ignition and power train software in automotive industry exceeds Six Sigma
Motivation
| |
| Page: 5 | Prev Top Next |
![]() Reality Gaps in Quality Processes
Many disconnects between what is taught and what
is done
Game companies ($9 billion dollar industry) do not market ISO certification ATMs do not meet Six Sigma access levels (5 min per year down) Open Source development a different paradigm Apache serve 70% of the web pages daily Firefox 25 million users in first 90 says CEOs managing products in new categories consistently fund new features over big fixing Each release moves further form Six Sigma Ironically, this is commonly the winning strategy
Motivation
| |
| Page: 6 | Prev Top Next |
![]() Measurements Matter
People are ruthless optimizers They will maximize results as per
Motivation
measurements If your measurements do not directly map to the desired outcome, then the results will be sub-optimal.
Do not ask staff to be martyrs
| |
| Page: 7 | Prev Top Next |
![]() Quality
Quality
| |
| Page: 8 | Prev Top Next |
![]() Economic Quality is
Usually a Brand vs Product measurement A multi-period measurement Expected future return
Quality
R
A time discounted measurement 1-d Vendor risk discounted measurement Less the probability cost of catastrophic
failure
R = R 1-d -
| |
| Page: 9 | Prev Top Next |
![]() Quality (Economic Quality)
Quality is delivering the user's expected economic return.
Quality
Quality= what we got so far expectations now return we initially expected
Q= =1
R 1-d
=1
n
=n1
n R 1-d
R 1-d
| |
| Page: 10 | Prev Top Next |
![]() Normal User Cost/Value Timeline
Value
Value
Value
Value
Value
Quality
Time
Cost Cost Cost Cost Cost Cost Cost
| |
| Page: 11 | Prev Top Next |
![]() Brand Value
H euristic: Do everything possible to make your word your bond. Being a reliable vendor has more impact to a user's expected return than promise of new capabilities.
Quality
Expected Return= what we think about the future
= R 1-d - R
| |
| Page: 12 | Prev Top Next |
![]() Keystone of UTSQ
Direct link between software quality and
firm's revenue. Maximum a Rational user will pay is capped by expected economic return.
Quality
Principle: Developer's revenue opportunity is capped by User's economic return - Quality.
max Rev= R u
users
| |
| Page: 13 | Prev Top Next |
![]() Risk of Catastrophic Failure
Task Value Phi Cost 3 Sigma Risk (1 in a thousand) Legal Documents $50 / day $1,000 $1.00 $0.001 $.999 Air Plane $500,000 / day $500 million $500,000 $500 $499,500 Nuclear Plant Man. $1 million / day $100 Billion $100 Million $100 Thousand $99.9 Million
Quality
6 Sigma Risk (1 in a million ) Gain
H euristic: H uge catastrophic risks dictate "heavy" development processes. Small dictate "response" processes.
| |
| Page: 14 | Prev Top Next |
![]() Heavy Processes
H euristic: As the ratio of catastrophic failure to period return increases, Six Sigma and similar processes closely map to Economic Return process .
Quality
Return= period returns- probability of the big one
R = R 1-d -
| |
| Page: 15 | Prev Top Next |
![]() Alternative Dev Processes
"First Article" Quality level X must be archived on initial launch. Large test suite. Focus on "corner cases."
"User Driven"
Quality
Quality level X by release Y. Test suite for mainline. "Corner cases" user reported.
Defect Density User Driven
First Article Time to Market
| |
| Page: 16 | Prev Top Next |
![]() "User in the Loop" Component of Theory
| |
| Page: 17 | Prev Top Next |
![]() Brand Value to User
Value
| |
| Page: 18 | Prev Top Next |
![]() Task Fixed Value
The work product for a Task has a value V that is independent of technology used to produce work product.
Technology A User Technology B Work Product T
| |
| Page: 19 | Prev Top Next |
![]() TaskWare
User in the Loop: Users are active components of the system not just passive consumers of functionality.
Tas k Wa r e Peo p l e Wa r e
Ta s k
Sof tWare
Va l u e
Costs
R=v T -c T
| |
| Page: 20 | Prev Top Next |
![]() PeopleWare
Ta s k Wa r e Peo p le Wa r e Sof tWare Pe o ple War e
R e s o u r c e Wa r e Kno wled geWare
KnowledgeWare: Ability to execute required actions unaided, i.e. The users have the know-how in their head. ResourceWare: Access external authority.
| |
| Page: 21 | Prev Top Next |
![]() Task Stimuli
c= f A t s f K t i f R t k
ts A ti K tk R
H euristic: To increase User return, systematically replace ResourceWare with KnowledgeWare and KnowledgeWare with software.
| |
| Page: 22 | Prev Top Next |
![]() Under Performing KnowledgeWare
. f1 f2 f3 f4 f5 ta . 20 . . . t b 200 . 90 10 . t c 75 . . . 5 td . . . 50 40
H euristic: Poor use of application is a Value killer. Work on understanding how skilled your users are.
| |
| Page: 23 | Prev Top Next |
![]() Parameters from Client Interface
User
Support
Software
Evaluate and Respond
| |
| Page: 24 | Prev Top Next |
![]() Summary
Summary
| |
| Page: 25 | Prev Top Next |
![]() Two Key Points
Probability-weighted cost of Catastrophic
Summary
Failure ( as measured by risk in UTSQ) is the best predictor of the development quality processes
Software Quality is measured by User's
Economic Return
| |
| Page: 26 | Prev Top Next |
![]() What does UTSQ Unify?
Unified Metric for: Direct link Quality and Revenue Features vs fix bugs resource allocation Help desk, documentation, training, UI research, ... resource allocation CEO decision domain and Engineering decision domain End user Economic Return is metric
H euristic: Invest in activities that create the max End User Return per dollar committed
Summary
| |
| Page: 27 | Prev Top Next |
![]() What Next
| |
| Page: 28 | Prev Top |
![]() Presentation and References
www.sqi-inc.com
| |




























