An Algorithm Decided I’m Not a “High Performer.” No One Can Tell Me Why.

We built a court that issues verdicts and has no appeals — and called it objective because it comes with a number.

Photo by Mohammad Rahmani on Unsplash

“The system flagged you as a developing performer,” my manager said, and then he looked slightly embarrassed, because he knew how that sounded and he knew he couldn’t explain it.

“Developing” is the word the company uses for the bottom band. There are four bands. I’d been in the second from the top for three years. Now an algorithm had moved me to the bottom, and the person delivering the news — my manager, a decent man who has known my work for years — could not tell me why.

I asked him directly. What changed? What’s the input? He pulled up a dashboard. There was a score, a number out of 100, and beneath it a set of category bars: collaboration, impact, consistency, growth. Mine had dropped in two of them. I asked what specifically drove the drop in, say, impact. He clicked around. The dashboard did not say. It showed the score; it did not show the reasoning. “It pulls from a lot of signals,” he said. “Calendar, project tools, comms patterns, peer interactions. It’s pretty sophisticated.” He said “sophisticated” the way you’d describe a weather system — something large and real and not yours to question.

So here is my situation. A system I cannot see has rendered a judgment I cannot interrogate, using inputs I cannot identify, weighted in ways no one will disclose, and that judgment now determines my raise, my standing, and — if it happens again next cycle — my job. And the human who is nominally responsible for evaluating me has quietly become the person who reads the verdict aloud.

I want to be fair about why companies do this, because the reasoning isn’t insane. Human performance reviews are famously terrible. They’re biased, inconsistent, driven by who the manager likes and who spoke up in the right meeting. Managers play favorites; recency dominates; the loud get rewarded and the quiet get overlooked. An algorithm, the pitch goes, looks at everything, evenly, without the politics. It’s supposed to be fairer. More objective. It sees the work, not the performance of the work.

And maybe it does see more. But here’s what nobody selling these systems wants to dwell on: removing the human bias didn’t remove the bias. It moved it somewhere I can’t reach. A biased manager is at least a bias I can see and argue with. I can know that Dave undervalues quiet people, and I can adjust, or push back, or go to Dave’s boss with specifics. The bias has a face and the face can be reasoned with. The algorithm’s bias — and it has one, because someone chose what to measure, and what you choose to measure is itself a value judgment dressed as math — has no face, no inbox, no capacity to be persuaded. You cannot take a dashboard to lunch and change its mind.

The thing that actually frightens me isn’t the bad score. Scores recover. It’s the disappearance of the conversation. Performance review, for all its flaws, used to be a thing that happened between two people who could look at each other and disagree. It was contestable. I could say “that’s not fair and here’s why,” and even if I lost, I’d been heard, and the person judging me had to hold the judgment in their own mouth and own it. Now the judgment comes from nowhere. My manager doesn’t own it; he’s just relaying it. The system doesn’t own it; it’s just software. The vendor doesn’t own it; they just built a tool, how you use it is up to you. Everyone in the chain can point upstream. The accountability has been distributed until it evaporated.

I asked the obvious question. Who can I appeal to? My manager said I could “add context for next cycle.” I asked who reads the context. He wasn’t sure. I asked whether the context goes into the system as a new input, or whether a human overrides the score based on it. He didn’t know. The honest shape of the answer was: you can talk, but it’s not clear anyone — or anything — is listening in a way that can change the outcome. The appeals process is a comment box that may or may not connect to anything.

We have built, inside the modern company, a court that issues verdicts and has no appeals. And we’ve done it while congratulating ourselves on objectivity, because the verdict comes with a number, and numbers feel like truth. But a number is only as objective as the choices behind it, and all the consequential choices — what to measure, what it means, how much each thing weighs — were made by people who are now nowhere to be found, encoded into a system that presents their judgment as fact.

What unsettles me most is how willing everyone is to defer to it. My manager, who knows my work is good, deferred to it. He believes the system sees something he can’t. Maybe it does. But notice what’s happened to him, too: he’s been demoted from a judge to a messenger, and he doesn’t seem to mind, because the system carries the weight he used to have to carry. It’s easier to deliver a verdict you didn’t make. He doesn’t have to defend it. He just has to read it.

I’m going to “add context for next cycle.” I’m going to do the thing the comment box asks. But I’m doing it the way you’d write a letter to an institution you don’t believe will read it — for the record, for my own dignity, not because I think it works.

And I keep coming back to the moment in that meeting when I asked who I could talk to about the decision. There was a pause. And then my manager turned his monitor toward me, a little apologetically, and gestured at the screen.

The answer to “who decided this about my life” was a dashboard. I was being invited to take it up with the dashboard. And the worst part is, I understood, in that moment, that there genuinely wasn’t anyone else to ask.