Every so often, when I think I understand everything that there is to know about a subject, I read a really interesting insight that causes me to totally rethink my assumptions. This is the case with a recent blog post by agile estimating and planning guru, Mike Cohn: How do story points relate to hours?
This is often a stumbling block for those new to agile estimation techniques that de-emphasize precise or padded time frames in favour of relative effort. After an iteration or two, some folks may begin to notice a correlation between points and hours, but this is a false-positive conclusion to reach.
Why? I’ll leave this crystal clear revelation to Cohn:
If the one-point stories are centered around a mean of x, ideally the two-point stories will be centered around a mean of 2x. This will never be exactly the case, of course, but a team that does a good job of estimating will be sufficiently close for reliable plans to be made from their estimates.
…[T]he relationship between points and hours is a distribution. One point equals a distribution with a mean of x and some standard deviation. The same is true, of course, for two-point stories, and so on…
A distribution is, of course, is a statistical term for the frequency or probability that variables take on in a sample. In this case, the variables are actual time records to finish (done-done) a one point story across a discrete time scale. Typical distributions, as they grow, tend to coalesce around a mean value, giving us a familiar graphical shape.
Cohn illustrates this with two distribution curves that will look familiar to most folks who survived high school math:
The overlap between the two distribution curves describes situations where teams’ estimates of a one point story on the high-end of effort converge with low-end two point stories. If you’ve observed how teams estimate using story points, this is an obvious (yet unstated) conclusion.
Excellent post – it totally revitalizes my interest in the topic of agile/iterative/lean project development.