2

I recall hearing some numbers related to the percentage time that is spent on analysis in a traditional software project. However, I'm now looking for references and coming up short. The only thing I have found so far refers to the Mythical Man Month where it states that 1/3rd of a project is spent on Planning. But I'm now looking for a more up-to-date reference. Does anyone have any other references from well known sources? TIA.

2 Answers 2

3

Nice question. The distribution of labour for software development across labour types varies per project, project approach and organisation. It has challenged me quite a lot in the past as an R&D manager. For current project we always work with some kind of FPA and in our case 15% for analysis unless reasons to deviate.

But that's not your question. Regarding references I've found most help in that period of my life with CMM (Capability Maturity Model, nowadays CMMi) and advice from Hans Sassenburg (nowadays living in CH). CMM induces that most organisations can not reliably predict the amount of analysis needed for a project till the variables influencing analysis time across the project have been brought into control. This includes also staffing ("replace the senior analyst by a junior which makes a lot of mistakes") and procedures ("waterfall" or "agile").

A nice quote from this presentation on predicting effort is ”You can’t fix what you refuse to measure”. This article on software cost prediction gives some more information.

Summarizing: either take a reasonable gamble based upon historical projects or ensure your projects, teams and customers are similar and more accurately calculate the amount of analysis time.

3

1) International Software Benchmarking Standards Group (ISBSG, http://www.isbsg.org/) collects this data, you can buy their report.

2) There also used to be (still is?) Chinese Software Benchmarking Standards Group (CSBSG) that collected this data for Chinese projects. I believe I saw their data some years ago, but I can't find it anymore, and their site www.csbsg.org is no longer up. I still see reference to them as a part of ISBSG at http://www.totalmetrics.com/function-points-groups/csbsg, for example.

3) I was able to find this report at COCOMO Forum'08 http://www.docstoc.com/docs/20160200/Phase-Distribution-of-Software-Development-Effort that compares CSBSG data with COCOMO II distribution - see slide 14 there. Acc to it, the Planning and Requirements phase is slightly above 15% for CSBSG, and slightly above 5% for COCOMO. Design phase takes another 15% acc to CSBSG and 35% acc to COCOMO model.

4) I also found another paper http://www.compaid.com/caiinternet/ezine/Reifer-Benchmarks.pdf (2004) that provides effort/duration distribution figures. In particular, the Plans and Requirements phase for waterfall is ~7% (2..15%) in terms of effort, 16-24% (2..30%) in terms of duration. Design phase is extra 17% / 24-28%. There are also distributions acc to other methodologies provided inside.

1

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.