technology supported requirement handling and estimation kjetil moløkken-Østvold – conceptos...
TRANSCRIPT
Technology Supported Requirement Handling and Estimation
Kjetil Moløkken-Østvold – Conceptos ConsultingNDC2010, 16.-18. June, Oslo
About the speaker: Kjetil Moløkken-Østvold
• Senior Partner at Conceptos Consulting• Academic and research background
– PostDoc and Assistant Director at Simula Research Laboratory (Norway)
– MSc (Siv.Ing.) and PhD from the University of Oslo (Norway)
– Published 23 papers on various topics, including estimation, project management, agile, collaboration, communication, contracts etc.
• Recent commercial projects– External quality assurance and consulting for several large
public sector projects (>500 MNOK)– Process improvement for financial and telecom companies
Today’s message
Good projectsBetter estimation,
requirements handling and process
Good collaboration
Contents
• Requirement handling and estimation• The case: Lindorff
– Problem– Solution
• Results• Conclusion
Requirement handling and estimation
Software Project Overruns
• About 70-80% of all projects encounter effort (cost) overruns¹
• The average magnitude of effort overruns is 30-40%
• Similar results for schedule overruns• No apparent change the past 30-40 years
¹ Moløkken-Østvold, Jørgensen, Tanilkan, Gallis, Lien and Hove. A Survey on Software Estimation in the Norwegian Industry, In 10th International Software Metrics Symposium (METRICS 2004)
Is requirement handling important?
• Requirement processes demand– Structure and a long-term perspective– Close collaboration with customers/users
• Should you have – A big up-front requirement process, – Only one for each iteration, or– A combination?
• Many have trouble with requirement processes in agile development
Requirement handling
• Changed and new requirements are perceived as the customers' most frequent contribution to overruns¹
• Overruns are prevented by the availability of competent customers and capable decision makers
• Avoid the influence of irrelevant and misleading information
¹ S. Grimstad, M. Jørgensen, and K. J. Moløkken-Østvold. The Clients' Impact on Effort Estimation Accuracy in Software Development Projects, In: 11th IEEE International Software Metrics Symposium (METRICS 2005), Como, Italy, September 19-22, pp. 3, IEEE, 0 ed.. 0, 2005.
Frequent communication
• Previous studies have found communication to be important for project success
• Frequent communication can be used to prioritize features, set focus on bug-fixing or include more functionality (Beck and Fowler, Planning Extreme Programming, 2001)
• Motivated in part by Cockburn¹, we explored the frequency of communication between the contractor and the customer
¹ Cockburn, "The End of Software Engineering and the Start of Economic-Cooperative Gaming," ComSIS, 2004.
Contact frequency and overruns
• A Kruskal-Wallis test for difference results in p=0.023• The corresponding size of effect is d=1.25, indicating a large
size of effect
Communication frequency
BREB
ias
Not dailyDaily
2,0
1,5
1,0
0,5
0,0
Boxplot of BREBias by Communication frequency Level Mean Median
Daily 0.09 0.19
Not Daily 0.58 0.35
The requirement pyramid ©
Pro
ject
man
ager
s
Dev
elop
ers
and
test
ers
Cus
tom
ers
End
-use
rs
Inte
ract
ion
desi
gner
s
Levels RolesCharacteristics
Pre
cisi
on
Com
mun
icat
ion
cost
s
Abs
trac
tion
Increases
Business goals
Requirements (coarse-grained)
User stories (etc.)
(Development) tasks
Functions (etc)
Agile development and estimation?
”-the concept of an overrun is not one typically found in agile development processes themselves, and precision estimation up front is not typically seen as a priority.”
- Comment from anonymous reviewer, Agile 2007, research track.
Main challenge
• It appears as if many books, papers and tutorials in the Agile community assumes:– A customer with no need for budget or
schedule when starting a project– No need for long term planning within the
development team(s)– That minor estimates derived as you go (e.g.
for sprints) are sufficient
Estimating Agile projects
• Agile projects are hardly immune to overruns, delays and bad business decisions based on poor estimates
• “Planning Poker” or similar techniques for estimating sprints or releases are important, but not sufficient
• It is often necessary to provide a relatively accurate estimate of total project delivery schedule and costs at en early stage, due to:– Bidding– Budgeting– Staffing– Scheduling– Release planning– All of the above…
Combination of estimates may reduce overruns
• Several studies have found that combination of estimates may reduce over-optimism and overruns¹
• True for both contract work and in-house development
• Combination in itself may be more important than the method chosen
¹ K. J. Moløkken-Østvold, N. C. Haugen, and H. C. Benestad. Using Planning Poker for Combining Expert Estimates in Software Projects , Journal of Systems and Software 81(12):2106-2117, 2008.
Why combine estimates?
• Obtain knowledge from various sources• Avoid extreme decisions• Synchronize perceptions about estimates and
work at hand• Create ownership of estimates• Remove irrelevant information (if using
moderator)• Introduce a ”devil’s advocate”
Pitfalls when combining estimates
• Passive participants• Depending on chosen method: political pressure
(groupthink)• Requires good moderators and experts• Time-consuming and costly (?)
Some methods for combining estimates
Method Structure Anonymity Interaction OverheadDelphi Heavy Yes No Major
Wideband Delphi Moderate Limited Limited Moderate
Planning Poker Light No Yes Limited
Unstructured groups Light No Yes Limited
Statistical groups Light Yes No Limited
Decision markets Heavy Yes No Moderate
Delivery (release) estimated with Wideband Delhi
• A release for NextLevel• Six estimators representing development and
business• Release was broken down to a set of requirements
(about 30 elements to estimate)Overall estimates Round 1 (hours) Round 2 (hours) Change (%)
Average (mean) 677 936 38,2 %
Median 577 991 71,9 %
Most optimistic 470 770 63,8 %
Most pessimistic 1027 1056 2,8 %
Estimation should be an integrated part of the requirement process
• Estimation, prioritization and planning of requirements provides new information
• Several roles should be involved in the estimation– Consider involving the customer as well
• Estimation should happen at several ”levels” (ref. the pyramid) in order to verify and triangulate
Project processes• A study found that effort overruns differed significantly
depending on project process¹– Sequential (waterfall): 55%– Flexible (agile, iterative): 24%
• Projects with a flexible process ranked higher on:– Good requirements– Good collaboration between developers and
business/customer
¹ Moløkken-Østvold and Jørgensen, "A Comparison of Software Project Overruns“, IEEE Transactions on Software Engineering, 2004.
The easy part when implementing agile processes
• Having some meetings while not sitting down
• Moving yellow notes around a whiteboard• Using strange job titles
The hard part when implementing agile processes
• Developing and prioritizing requirements• Providing (accurate) estimates• Collaborating between developers and
customers
The hard part is even harder in a distributed
environment
The case: Lindorff
• Lindorff Group is a leading outsourced receivables management company in Europe and on a global basis
• Lindorff has approximately 2200 employees • Offices in Denmark, Estonia, Finland, Latvia, Lithuania,
Germany, the Netherlands, Norway, Spain, Russia and Sweden
NextLevel Project
• Purpose of project– To deliver a new .NET GUI on an existing debt
collection application– Re-write business logic from Powerhouse to
PL/SQL and .NET
NextLevel Project
• Distributed participants– Developers in Oslo and Bø– Testers and business developers in Oslo,
Røyken and Trondheim– Users/customers in Røyken, Oslo, Trondheim
and other locations
Internal improvement study Q4 – 2007, findings:
• Poor routines for communicating and prioritizing requirements
• Effort overruns (about 25%)
• Internal collaboration rated as average
Describing, refining, prioritizing and communicating requirements is difficult for business people
Understanding, estimating and developing requirements is difficult for software developers
Problem: inefficient communication
Solution
• Implement a flexible sharing regime for requirements, prioritization and estimates
• Lindorff joined a project financed by Innovation Norway
• Main development partner was Symphonical
About Symphonical
• Symphonical is a webtool that takes you from idea to action
• It incorporates document handling, group collaboration and simple task tracking using digital post-it notes (information containers) on walls
• Cloud based (Amazon) since 2008
• Spin-off from Simula Research Laboratory
Example features used by Lindorff
Live
• Allows for synchronized collaborative work– Actions performed by one person is shown in
real-time to others logged in• Used in
– Requirement workshops– Requirement reviews
• Used when the meeting is held in more than one location
Sharing
• Share one or more information containers (notes) with collaborators
• Used in– Requirement specification– Requirement quality assurance
• Used to get input from resources outside the project– Sporadic contributors
Voting/estimation
• Used for estimating requirements/user stories in collaboration– Used in advance of the
developers estimation meeting, i.e. the first estimation iteration
– Used in the estimation meeting to administrate estimation iteration 2 (and if needed; iteration 3)
• Can also be used to prioritize requirements etc.
Voting/estimation (cont.)
Conversations
• A history of in-context discussions, replaces mail etc.
• Used through out the requirement process, from the initial workshop to the final quality assurance
• Helps keep track of the evolution of the requirements
Template
• A default text on notes (requirements/user stories)
• Used– in order to make sure the requirements are
expressed in a format that makes sense to the developers
– to ”force” the requirement handlers to provide mandatory information
Tool
• Reuse settings/templates• Used in order to be able to reuse
structures and default text on notes in later projects/deliveries
• Supports the use of the chosen methodology
Results
• Lindorff improved on all studied areas• Respondents provided ratings (1-5, 1=best)
Main results (two years after)
Selected areas 2007 2009
Collaboration business/IT 2,4 1,9
Requirement handling 3,0 1,9
Estimation process 3,3 2,2
• Effort overrun (perceived) was reduced from an average of 25% to 12%
Responses from Lindorff
• Requirement handlingo All respondents (IT and business) reported that use of
Symphonical had greatly improved how requirements were specified, refined and communicated
• Estimationo About half reported that Symphonical had improved the
estimation processo The other half reported no discernible impact from
Symphonical by itself (though it could have played a role along with other improvements)
o Responses to a large extend depended on company role, with business perceiving most value
Respondent quotes
• “I perceive [the change to Symphonical] as positive [...] better than Excel [...] having every requirement as a use case is good. Structure, retrievability and traceability is much better than in Word and Excel.”
• “It is easier to get input from more people now [...] it is easier to concentrate on one single requirement at a time.”
• ”…Previously, one studied and estimated in meetings. Now it is possible to prepare, estimate and think [...] one can view average values and there is traceability.”
Conclusion
Lessons learned
• Analyze your situation regarding quality of requirements and magnitude of effort overruns
• Get rid of static documents (Word, Excel etc.) shared via email and informal communication (calls, notes on desk etc.)
• Implement a web-based collaboration platform
The future of work?
Q&A
Thank you!
• Contact information:– Presentation: http://www.conceptos.no/– Email:– Please visit our stand here at NDC!
• This research project was– Funded by Innovation Norway– Conducted in collaboration with Dr. Kjetil Kristensen, founder
and Principal Consultant of Kristensen Consulting
• Caveat Emptor: the author of this presentation is a member of the board of directors at Symphonical, and has ownership interests in the company