Velocity is a range
An average velocity produces plans missed half the time. This gives you the spread your team actually delivers within, and forecasts from that.
Velocity as a range, not a number
An average velocity produces plans that are missed half the time. The spread is the useful output.
Sprint by sprint
What that means for planning
How this is calculated
Mean, median and sample standard deviation (n−1) across the sprints given. The planning range is the 15th to 85th percentile of what actually happened, not a multiple of the standard deviation — velocity samples are small and rarely normal, so reading percentiles off the real data is more honest than fitting a curve to eight points.
Stability is the coefficient of variation: steady under 0.15, variable under 0.3, erratic above. It describes consistency, not quality — a steadily low velocity is steady.
The trend is a least-squares slope, shown only with four or more sprints and hedged when shown. Four points can look convincingly linear by chance.
Velocity is not productivity
It is a property of one team's estimating scale. Two teams delivering identical work can differ threefold in velocity purely by how they size things, and a team that inflates its estimates has a rising velocity and no more output.
This makes velocity useless for comparing teams and actively harmful as a target. The moment it becomes something to increase, it increases — through estimation, not delivery. The number stops measuring anything and the team loses its forecasting tool at the same time.
Plan with the range
A commitment built on average velocity is missed about half the time, which is not a failure of the team so much as the definition of an average. If six sprints ran between 18 and 26, then a six-sprint plan is somewhere between roughly 110 and 155 — and quoting the middle of that as a date discards everything the data told you.
When velocity stops working
Velocity assumes a stable team, a stable definition of done and a stable estimating scale. Change any of them — two people join, the definition tightens, the team recalibrates after a painful sprint — and the history before the change is describing a different team.
When that happens the honest move is to start the series again rather than average across the discontinuity. Three sprints of relevant history beat ten sprints of history about a team that no longer exists.