Skip to content
Delivery & Execution

Velocity Is a Planning Instrument, Never a Performance Metric

A director asks why one team's velocity is 41 and another's is 23. The question is wrong, and the engineering manager is the person who has to say so, and offer something better.

Delivery & ExecutionEngineering Management Academy5 min read

A director looks at two sprint reports and asks why Team A's velocity is 41 and Team B's is 23. It sounds like a reasonable question. It is the question that turns a planning tool into a weapon, and the engineering manager is the person standing between the two.

We think velocity has one legitimate use, and it belongs to the team that produces it. Everything else people do with it, and the list is long, makes the number worse at its one job.

What velocity is for

Atlassian's guide to velocity in Scrum describes the mechanics well. At the end of a sprint you add up the story points for every story that was fully completed, and only those. A story that is ninety per cent done scores zero. Average the last few sprints and you have a figure the team can plan the next sprint against: if the average is 15, commit to roughly 15.

Read the same guide to its end and it says the two things managers most often forget. Velocity "is not the same as productivity", and "it's not a measure for comparing the performance of different teams." A lower number does not mean a worse team. Each team points its own work in its own unit.

The Scrum Guide is quieter still. It never mentions velocity or story points. What it says about Sprint Planning is that the more the Developers know about their past performance, their upcoming capacity and their Definition of Done, the more confident they will be in their Sprint forecasts. Past performance is an input to a forecast the Developers make. Nothing in the framework hands that number to anyone above them. On tracking tools generally, burn-downs and the rest, it adds that they do not replace the importance of empiricism.

So the framework's answer to the director is that the number was never addressed to them.

How the misuse begins

Ron Jeffries, who was in the room when story points were first used, wrote in Story Points Revisited that he may have invented them and, if he did, he is sorry now. His regret is about what happened to them. Points began as ideal days renamed so that stakeholders would stop reading them as calendar days. Then organisations started using them to compare teams, assuming the team with the lower number was the weaker one rather than asking what differed in its work. Then they demanded actuals against estimates, to check the estimating, which pulls attention away from delivering. And then the pressure for more, which teams meet by thinning testing and quality, which raises the defect rate, which slows them down.

Atlassian's own guide shows how easily this slides. In one paragraph it offers a benchmark of five to ten points per person per fortnight for new teams. In another it suggests that a velocity climbing from 40 to 60 shows that process improvements are paying off. A per-person benchmark invites exactly the comparison the FAQ later warns against. And a velocity that rises by half with the same people and the same codebase has, in our experience, almost always been re-pointed rather than improved. Ask any team that has been told to raise its velocity by twenty per cent. It took one sprint.

Three rules the manager enforces

The manager cannot stop the director from asking. The manager decides what happens next.

Velocity stays inside the team. It appears in planning and in the retrospective, and nowhere else. It is not in the monthly slide, the quarterly review or the one-to-one. The moment it leaves the room it becomes a score, and a score gets gamed by honest people who are trying to look like they are doing well.

Teams are never compared on it. When the director asks, the answer is that points are a local unit, like a team's private currency, and that comparing 41 to 23 is comparing dollars to yen without a rate. Then offer something they can use: a dated forecast with a range, and the list of what each team shipped. That takes two minutes to prepare and answers the question they were trying to ask.

No velocity target, ever. A target on velocity is a target on the unit of measure. Jeffries' description of the pressure for more is the predictable result. If you want more throughput, change something real: the dependency that stalls every sprint, the review queue, the on-call load. Then watch the forecast, not the points.

When the number moves

A drop in velocity is a question, never a verdict. Atlassian lists the usual causes and they are mundane: a dependency on another team that ran late, holidays, a sprint loaded with harder stories than usual. We would add two from the manager's side of the table. An on-call rota that puts two of five engineers on incidents for a week halves the sprint and no chart will say why. A pull request that waited two days for review because the only reviewer was in discovery interviews shows up as an unfinished story, which scores zero, which looks like a slow team.

So when the number drops, go and look. Count the stories that were started and not finished, ask what each one was waiting on, and you will usually find the cause in a day. When the number rises, be more suspicious, not less. Ask whether the stories got smaller or the points got bigger.

Jeffries' alternative is worth taking seriously even if you keep points. Slice stories until each one is a day or two of work, and count stories instead. Small stories make counting and pointing converge, and a count is much harder to inflate than an estimate. Whatever unit you keep, the rule holds: it is for the team to plan with, and for nobody to be judged by.

Where this sits in the standard

Honest planning sits in the Delivery & Execution domain of the EMA Competency Framework. The Body of Knowledge unit on estimation, planning, and honest commitments covers what a forecast owes a stakeholder, and engineering metrics and measurement covers why a measure that leaves the team that produced it stops measuring anything.

In EMA-I it comes up as a situation: a stakeholder asks for a number, and you decide which one to give them.

Filed undervelocityscrumstory pointsestimationengineering metrics
This sits under Delivery & Execution in the Competency FrameworkThe Body of Knowledge chapter covers the full set of competencies the examinations assess in this domain.Read it →

Test your judgment, not your recall.

EMA-I is a scenario-based examination in the five domains of the framework. There is no course to buy and no attendance to prove. Sit a sample first, or read what the credential covers.

Keep reading