Two folks asked me to write a bit more about the afternoon sessions, especially about the estimation one and here I go, I will resume what we talk in that session.
The format of the session was open, there were more that 20 people and each talk about the way he uses to estimate.
In may opinion the natural process is to first estimate the size, then the effort, and finally considering a certain team you can deduce duration and cost, but not everybody uses this process (or unless not explicitly), some stated to estimate duration at once.
In few words, the following methods were mentioned:
- Parametric methods (or magic formula-based methods that is the way I like to call them), basically this methods propose to analize your requirements and count points to determinate the size. Then you use that points in a formula that adds some variables to adjust the estimation based on particular information of the project (programming language, experience in similar projects, etc). In this category are Cocomo methods.
- Expert opinion and manager manipulation methods (this original name is mine): maybe this is one of the most commonly used methods. Several people in the session told to used it and I know that some colleagues in my company use it all the time. It is very simple, a manager (boss, project leader, project manager or whatever you call it) ask an expert (some time not so expert) to estimate certain items. Then the manager adds some buffer and that’s it. Not very formal, not very precise but fast and easy.
- Several-experts-based methods: I know this is a very ambiguous name, but that is the idea, I want to use this name to refer several different methods that include Delphi methods and planning poker (I think that planning poker is a Delphi variant). In these methods you have a team of experts (when I say experts I am just referring to people that is capable of give a realistic estimation because of his experience). Something very interesting of these methods is that under certain conditions these methods get a probabilistic support: if there are 4 or more experts and each expert opinion is independent, then we can apply the central limit theorem and get an confidence interval.
In my case I always try to use methods in the third group, in particular to estimate the complete product backlog at the start of a project I prefer to use Wideband delphi and then when estimating tasks at the begining of an iteration I prefer planning poker. During the session I mention a spreadsheet I use for Wideband-Delphi and after the session some people asked me to share the spreadsheet, so I published it here.
When I started writing this post I get the idea of preparing an estimation workshop, I think it could be an interesting activity, well I will analyze it.
That ‘s all, see you.