uksma 2005 project estimating – theory and practice – beyond talking paul goodman
TRANSCRIPT
UKSMA 2005
Project Estimating – Theory and Practice – Beyond Talking
Paul Goodman
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 2
consulting
Our Starting Point
“Our customers quite rightly complain.
It is not the fact that we get it (our estimates) wrong!
It is that we are always late, never early!”CIO’s recent comment on the state of estimating and the effect on customer satisfaction
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 3
consulting
Topics for Discussion
Why is estimating accurately important?
Where should we start our improvement from?
A Case Study Overview
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 4
consulting
Estimation accuracy is a critical success factor for most businesses
Getting it wrong can adversely affect your company bottom line
And Will certainly upset your customers
Basic principles:
The Importance of Accurate Estimating
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 5
consulting
The Importance of Accurate Estimating
We should not expect accurate estimating to be “natural” i.e. to just happen
But It is possible to estimate, after requirements specification, to within 5% of actuals
Accepting 200%+ over runs is not the way it has to be
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 6
consulting
Starting Improvement
Appreciate that this will not happen overnight
It will probably take at least one year elapsed to effect broad sustainable improvement
Get to grips with some basic principles
Some are facts of life but they still need to be clearly recognised and stated
Others need to be embodied within your estimating processes
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 7
consulting
Facts of Estimating Life
Estimation inaccuracy is the difference between the estimate and the actual
If reducing this difference is important to you, make it important to everyone with estimating responsibility
If it is not important to you – STOP NOW!
Estimates and actuals need to be recorded in order to enable and to show improvement
Appreciate that people are genuinely and almost universally optimistic
Optimism is dangerous when estimating!
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 8
consulting
The estimator has the last word Estimates are estimates - use bounds and a most
likely The estimate is dynamic - re-estimate Use a sanity check, more than one technique Keep information
Estimating Principles
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 9
consulting
Recording Information
InitialEstimation
Re-estimation
Size ChangeControl
FirstEstimate
RevisedEstimates
ProjectCompletion
Actuals
InitialSize
RevisedSize
ProjectCompletion
FinalSize
Metrics Database
Identify the WEDGE factor
Effort
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 10
consulting
The Wedge Factor
base line requirements
final requirements
70% growth!
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 11
consulting
Mark II Function Point Analysis IFPUG Function Point Analysis Work Breakdown Structure/Task Sizing Line of Code Sizing for Infrastructure Projects
Effective sizing at an early stage is a pre-requisite to accurate estimation
Common Sizing Techniques:
The Need for Sizing Requirements
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 12
consulting
Using Estimation Tools
There are many estimating tools available on the market
They can be very useful
They do need to be used with care
Be aware of the need to calibrate
x
For internal use of <FullCompanyName> only. © 2003 Gartner, Inc. and/or Gartner Holdings Ireland.
All Rights Reserved.Page 13
FullCompanyNamePresentationTitleEngagement: xxxxxxxxx—DayMonthYear
consulting
Estimating Case Study
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 14
consulting
Agenda
Introduction and Process
Sizing
Costing
Effort and timescales
Summary
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 15
consulting
Introduction
Meta Group UK was requested by XXX to provide size estimates for the YYY System (a GIS application)
META Group were further asked to provide indicative and comparative costs estimates based on industry benchmarks
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 16
consulting
Conduct of Assignment
Chosen size metric was Mark II Function Point Index (FPI)
This was to be estimated based on best available information provided as documentation by XXX
Clarification of issues on documentation was provided by XXX
Documentation was reasonable and facilitated the use of a number of FPA estimation approaches
(Fact of life – the more limited your information, the less able you are to estimate)
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 17
consulting
Lifecycle Coverage
Our estimates will include the following elements of a project: Requirements
Design
Build
Integrate
Test
Project Management
Quality Assurance
Excludes effort associated with: Architecture and infrastructure (plan, build or run)
Business input
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 18
consulting
Cost/Effort estimate
Function Points are an estimate of size, not cost
The cost of delivering a system will depend on many factors including:
Maturity of the development teams
Stability of requirements
Tools and processes used
Quality of code developed (speed of test/acceptance)
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 19
consulting
Findings
Size estimation using three techniques:
Structured Expert Judgement; (Confidence is +/- 40% but tends to underscore)
Data Model Approach; (Confidence is +/- 25%)
Functional Decomposition (calibrated through Data Model Approach); (Confidence is +/- 20%)
(These are in the public domain – see “Sizing and Estimating Software in Practice”, Treble and Douglas)
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 20
consulting
Size Estimation Results
Technique Lower Est Most Likely Upper Est Bounds
Expert Judgement 1,600 FPts 2,700 FPts 3,800 FPts +/- 40%
Data Model 2,500 FPts 3,300 FPts 4,100 FPts +/- 25%
Functional Decomposition
3,000 FPts 3,800 FPts 4,500 FPts +/- 20%
META recommendation is that most likely size estimate be taken as 3,500 FPts
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 21
consulting
Sizing by Contribution
Contribution to total size made by Enquiry/Report functionality appears low
18% of total business functions are Enquiry/Report type contributing 12% of total functionality
This may be perfectly valid but raises a flag – “Is XXX confident that reporting functionality has been adequately identified?”
It is stressed that this observation is based only on estimated counts.
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 22
consulting
Cost Estimate 1
Assumptions:
System size is 3500 Function Points
Average lifecycle delivery rate of 15 FP / Man Month (mid-point performance within META Group’s AD benchmark)
Note: GIS type systems tend to achieve lower productivity levels than some other system classes
Blended ‘daily rate’ of £650 per day (mid-point in META Group’s Labour Rate benchmark)
Total Function Points
3500
Total months of Effort
233
Cost per ‘man month’
£14,300
Total cost £3,340,000
Note: Based on 22 working days per month
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 23
consulting
Cost Estimate 2
Assumptions: System size is 3500 Function Points
Based on International Software Benchmark Standards Group median productivity of 0.14 FP per Person Hour
Blended ‘daily rate’ of £650 per day (mid-point in META Group’s Labour Rate benchmark
This is an ‘optimistic’ estimate as this benchmark tends to reflects ‘best in class’ development productivity
Total Function Points
3500
Total Days of Effort 4200
Cost per Day £650
Total cost £2,730,000
Note: Based on 22 working days per month and 6 effective hours per day
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 24
consulting
Cost Estimate 3
Assumptions: System size is 3500 Function Points Based on CCTA median productivity of 0.07 FP
per Person Hour Blended ‘daily rate’ of £650 per day (mid-point in
META Group’s Labour Rate benchmark This is an a more pessimistic estimate based on a
less mature development environment
Total Function Points
3500
Total Days of Effort 8300
Cost per Day £650
Total cost £5,400,000
Note: Based on 22 working days per month and 6 effective hours per day
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 25
consulting
Costing Summary
META Group propose that the yyy application, based on documentation provided, is approximately 3500 FP in size ( with a likely accuracy of +/- 20%)
All evidence suggests that the cost range is between £2.7m and £5.4m based on market blended rate of £650 per day
META Group have no appreciation of the development maturity of the potential suppliers but do note that this has a major impact on costs
Cost may be reduced by the use of pre-packaged software components and the effective use of code generation tools (if applicable)
Overall, our experience indicates that based on our labour rate the total cost is likely to be between £3.2m and £4.5m
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 26
consulting
Reality Check
At this point we were told that the latest in house estimate was in the area of £3M
Given the discussion about reporting functionality that was now considered “low”
So what about the duration estimates?
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 27
consulting
Delivery Timing
We looked at duration estimates based on the estimated size and the profile of the organisation/project (mainly using COCOMO 2000)
The indications were that delivery within a 12 month period (including test) is an extremely aggressive timescale
This is putting it politely
This was in line with the “gut feel” of the teams on supplier and client sides (but now they felt able to talk about it)
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 28
consulting
Summary
Our best estimate is that the application should cost between £3.2 and £4.5m based on market rates
Delivery in a 12 month period would require a team of 15-23 FTE (and considerable “good fortune”)
The supplier would need to focus strongly on its development performance to achieve these targets
We recommend the supplier use function point techniques through the process to assure XXX of best in class development performance and to control change
© 2003 Gartner, Inc. and/or Gartner Holdings Ireland. All Rights Reserved.
Page 29
consulting
Conclusions
This exercise took approximately two weeks elapsed time
The approaches described and data used are mostly in the public domain (or available at relatively low cost)
The analysis provided an objective vehicle for discussion
Estimates were revised towards the upper bounds and these revisions were informed by the analysis
Duration estimates were relaxed somewhat
Nobody discounts the need to apply some judgement
The project is on track to deliver within budget