ProjectVelocityQuestions

I worry when I read statements like "we're going twice as fast ..." based on ProjectVelocity measurements because velocity depends on level of optimism in the team when the estimates are made.

If I want to compare one iteration with another a year ago and I know the team is made up of new or other team members am I just kidding myself when I see improvements? maybe everyone just got better at estimating (more pessimistic)? RachelD note we deviate in our team from DoingXpByTheBook by estimating as a group rather than as individual developers?

Maybe you're kidding yourself, and maybe you aren't. Take a card from now, and a card from then, and put them side-by-side. Do they seem to represent the same amount of complexity? If so, then the improvement may be real, if not, not. Hopefull everyone has got better at estimating (more accurate). There's a good chance that the current iteration involves cards estimated recently and cards estimated some time ago, so there's a levelling involved there.

Writing as an unrepentant rate doubler, let me tell you what our graphs show. They show a curve with significant wobbles. A least-squares fit shows a definite upwards trend. That could mean any one or more of a lot of things. More significant is a moving average (with window of three or four iterations). This shows a remarkably steady (but low) mean velocity for maybe the first third of the project, then almost a step change (over a few iterations) to a different mode, where the mean is much higher (more than doubled), but the variation is also greater. This step doesn't seem to correspond to any particular external event in the project history. But something happened. Just from the graphs it's not clear whether that was us suddenly getting much better at estimating, or suddenly getting much better at working on the project.

Combining knowledge of the project internals with the graphs suggests strongly that it wasn't an improvment in estimating, since there wasn't a lot of estimating going on at that time. Having been there, what it does correspond to is us getting our act together, particularly with respect to the build process, and if I remember correctly, it's about the time we started running acceptance tests in ernest. That said, we did get better at estimating over the lifetime of the project, and felt no compunction about re-estimating upcoming cards that seemed, with our greater experience, wildy wrong.

Also, although I don't have any hard evidence to back this up, I have a strong impression that we over estimate about as often as we underestimate.

RachelD I understand the basis of your statement better now. However, the experience in our team has been that we make no consistent velocity improvement. This probably has a lot to do with the small size of stories and the rate of change of the product requirements (the developer team has not had significant changes). We seem to get a boom-bust effect (I'll post some examples soon). ³host³³date³April 18, 2002³agent³Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 5.0)³ProjectVelocityQuestions