Stop Measuring Velocity. Start Measuring How Often You Were Right.

I watched a project run a beautiful few weeks by the numbers. Throughput up, everything moving, the burn-down pointing the right way. And it was quietly heading somewhere slightly wrong the whole time, faster than it ever could have before. By the time the direction problem surfaced, we were further down the wrong road than the old pace would ever have let us get.

The chart was green throughout. That was the problem, not the reassurance. Every number I report on said the project was healthy, and every number was true, and the project was not healthy. That gap is the whole point.

I do not blame anyone on that team. They were doing good work, quickly, and every signal they had said they were winning. That is the trap in one sentence. The metric they trusted was not lying about the past, it was accurately reporting motion, and motion was exactly the thing that had quietly stopped meaning progress.

Velocity was always a rough proxy. More shipped, more done, roughly the right way. It worked because, when building was slow, output and progress travelled together. You could not produce much in a wrong direction before the effort of producing it made you stop and look.

Now you can. Output went up sharply and the certainty that it was the right output did not move at all. So the two things velocity quietly bundled together, how much you did and whether it was worth doing, have come apart. Measuring the first now tells you almost nothing about the second, and measuring it well can actively mislead you, because a high number feels like health.

It is worse than useless when it is high, actually, because a high number buys the wrong direction more time. Everyone looks at the chart, sees green, and does not go looking for the problem, precisely when the speed means the problem is compounding fastest. The metric that was supposed to warn you has become the thing reassuring you while it burns.

A softer, more embarrassing metric: how often I was right. Not shipped, right. When I said "on track," was it. When I flagged a risk, was it the risk that mattered. When I chose to leave something out as noise, was it noise. When I decided what to build, was it the thing worth building.

None of that fits on a dashboard, which is exactly why it is honest. You can only judge it after the fact, by keeping a quiet count of your own calls and how they turned out. It is uncomfortable, because it makes you the metric instead of the team's output. But it measures the thing that actually became scarce, good judgment, rather than the thing that became cheap, motion.

There is no clean dashboard for this, and I have made peace with that. Some of the most important things are only visible in hindsight, and only to the person who made the call. Writing them down honestly, even when the honest answer is that I was wrong, is the closest thing I have found to a real measure of whether the job is being done well.

I am not saying throw out velocity entirely. It still tells you something about capacity and flow. I am saying stop treating it as a health signal, because a fast team pointed slightly wrong will show you the best numbers of its life right up until the rework lands.

The habit I am trying to build is small: at the end of a stretch, before I celebrate what shipped, I look back at the handful of judgment calls I made and ask honestly how they turned out. It does not feel like measurement. It feels like keeping score against myself, which is roughly what it is.

It will never make a status report, and it will never go on a wall. But it is the one that would actually have caught the green-chart project before the rework landed, because it asks the question the dashboard cannot: not are we moving, but are we moving toward the right thing. Everything else is just speed, and speed was never the thing in short supply.

So if your charts look great this quarter, here is the question I would sit with: are you tracking how much you shipped, or how often you were right, and when did you last check the gap between the two?