The team estimated stories in points: one, two, three, five, eight.
A junior asked, “What is a point?”
The senior developer replied, “A point is what we agree it is.”
“That seems arbitrary.”
“It is. That is why it works. The moment we measure it precisely, we will be measuring the wrong thing.”
Beyond the Surface
The junior is right. A story point is arbitrary. The senior’s move is to agree completely and then explain why the arbitrariness is load bearing.
Points survive in software teams because they are deliberately imprecise. A single number holds uncertainty, relative size, unfamiliarity with the code, and the shared intuition of everyone in the room, and it holds all of that without ever becoming a promise. When two engineers argue about whether a story is a 3 or a 5, the number is not the point of the argument. The argument is the point. One of them knows something about that part of the system that the other does not, and the estimate is the excuse that surfaces it.
Trouble arrives the moment someone converts a point into hours. Now it is a unit of time. A unit of time can be compared across teams, tracked quarter over quarter, and put on a slide in front of people who have never read the codebase. The estimate has become a target, and an estimate stops being honest the moment it can be held against the person giving it. Every team that has been shown its own velocity chart in a performance conversation learns to inflate points within one sprint. The measurement did not become more accurate. It became less true.
This is the older lesson underneath the koan, the one economists named after Goodhart. A measure that becomes a target ceases to be a good measure. Story points are unusually vulnerable to it because they look quantitative, they sit in a tool that graphs them automatically, and nobody outside the team can tell the difference between a team getting faster and a team getting more generous.
So sit with the question the senior is really asking. What do you want to know when you ask for an estimate? If you want to know whether the team understands the work, points do that well. If you want to know whether a date is safe, points have never been able to tell you, and making them more precise will not change that. The precision would only give you a more confident way to be wrong.