The sprint velocity is a mean

One of my previous teams would give a five-finger confidence rating after each planning session to see how each team member felt about the plan. Like most times before, the median was a 4. After all, we had brought in the same number of stories as the velocity. Everything sounds good, right?

After playing around with the velocity forecaster, something hit me that I hadn’t really considered before, even though I knew that we use the average of the last couple of iterations as our velocity. Since velocity is a mean (assuming the central limit theorem holds true), we should hit that value 50% of the time.

To put that in context, let’s look at some of the previous points delivered on a project:

36, 63, 21, 93, 63, 63, 77, 63, 22, 65, 31

The average would be 54.3 (90% CI: 41.6-66.9). Out of this sample, 63% are above the average, which seems ordinary for a small sample. This means the mean is holding relatively true. To test this more concretely, we would need at least 30 data points.

This post is pointing out quite an obvious fact in hindsight, something I hadn’t really considered. In some planning sessions, the sprint velocity and the goals based on it are often discussed as sure bets, something we can be confident showing the client. In reality, it’s a coin toss.