A word written about a person is a claim, though almost nobody writing one thinks of it that way. You put it in a form in March because the form asked for one, and the form is honest about what it wants: a single word that can be sorted and compared and carried into a room where nobody who works beside that person is sitting. The word was true the day it was written. What makes it a claim rather than a record is everything that comes after, because from the moment it is filed it gets read as though it still describes somebody.
A claim about a person, drawn from what somebody saw, is worth exactly as much as the next look proves it worth. That is not a criticism of the claim. It is what a claim is. The obligation it carries is the ordinary one: you commit in advance to checking it against what happens next, on a cadence set before you know the answer, and you let the answer come back different. A reading somebody keeps checking is a reading. A reading nobody checks again is an opinion that was filed.
Where this goes wrong is not where anybody expects. Attention is finite and it flows toward bad news, so the readings that come out poorly get read again within the month, and the readings that come out well are the ones that stop being read at all. The company has quietly decided that a strong word is a conclusion and a weak word is an open question. It is the wrong way round, or rather it is a distinction the method does not support. A high rating is a status like any other status, and it requires measurement like any other status. Not heavier measurement, and not lighter. The same instrument, on the same return, asking the same few questions.
This is worth saying carefully, because the moment you talk about continuing to look at somebody who is doing well, it sounds like suspicion, and suspicion is the opposite of what is meant. Nobody is on probation. The point is that the ordinary process everybody else stays inside should not have an exit at the top of the scale. Today it does, and the exemption is invisible because it looks like trust and it feels like leaving a good person alone.
So take John, who had a genuinely strong quarter. He carried the payments migration further than anyone expected, he reviewed more than he was asked to, and on a Thursday in March his manager told him so and meant it. Good feedback, honestly given, landing on work that honestly earned it. Nothing in that conversation needs fixing. The only thing worth noting is that it was the last observation anybody made.
Because the reading came out good, nobody read again. By the second week of April John was alone on the migration, the two colleagues who knew it having moved to the thing that was on fire. By May his reviews were arriving slowly, then barely, which is the kind of change that never announces itself. In June somebody asked how he was doing and the answer came back from March, because March was the last time anybody looked.
Nothing here was a failure of judgement. The March reading was accurate. It was simply never tested, and an untested claim does not stay accurate on its own.
Which is why the thing worth caring about is the instrument rather than the status it produces. Companies maintain the status and neglect the instrument. They rewrite the rating scale, they calibrate the labels across teams, they argue about whether there should be four levels or five. All of that is work on the output. Almost none of it is work on the thing that captures the information the output is made from, and the output cannot be better than what fed it.
The instrument does not need to be elaborate. It needs a short memory and a reliable return. The short memory keeps the reading about now, because evidence that piles up from the day somebody joined stops being a reading and becomes a dossier, and a dossier is what makes a person defend March in June. The return is the promise that somebody is coming back to ask the same few questions and to let the answers be completely different, because the month they are about is a different month.
The meeting a manager holds with each person is already this instrument, running badly and by accident. The manager arrives with a picture of the last fortnight, assembled out of what they happened to see, and the person in the chair is the only one who actually lived it. The distance between those two is the useful thing in the room. It corrects the picture, and where anybody is paying attention it corrects the way the next picture gets made, because a manager who was wrong about what mattered in April has just learned what to watch for in May.
What such an instrument surfaces is usually absence rather than fault. In April it would have said John’s name, and what it should have said is the hardest thing to say without being misunderstood: nothing was wrong. No failure, no missed deadline, nobody complaining. There was less of everything than at the start of the month, fewer of the small ordinary signs, and that thinning is the exact shape a month takes when somebody is holding something difficult alone. Said as a finding about John, it lands as an accusation and the room closes before it opens. Said as an instruction about where to point the next hour, it is only a question you expect to be wrong about. Either he says the migration is fine and means it, and the signal was noise, which is a result. Or he says the two people who knew it are gone and he has been carrying the reviews alone, which is April, found in April.
None of this argues that the verdict should go away. A company has to decide who is promoted and who is paid what, somebody has to be accountable for saying it out loud, and a team kept under permanent observation with no call ever made is being watched rather than cared for, by a manager avoiding the hard afternoon. A person deserves to be told where they stand, in a word, on a date, by somebody willing to defend it.
What the cycle does not earn is the months between. The word from March was a photograph taken in good light, and the mistake was never in taking it. The mistake was hanging it on the wall and walking past it until October, as though a person were a thing that could be finished. A photograph is a moment and it is worth having. The film is what actually happened, and a status only means something as one frame of it. The good news in March was true. All it ever asked of anybody was to come back in April and find out what it had turned into.