Performance management without the theatre

I don't run a formal performance review cycle. No annual, no semi-annual, no ratings, no self-assessment form that takes an engineer four hours and gets read for ninety seconds.

That reads as a soft position and it's the opposite. Removing the cycle raises the load on managers rather than lowering it, because the cycle's real function in most companies is to be the one time of year when the conversation is guaranteed to happen. Delete it and you've deleted the guarantee, so the feedback has to actually be happening in the other fifty weeks or the whole thing collapses into nobody knowing where they stand.

Which is the trade I'll take. The alternative is a process that produces a document nobody acts on, a week of lost engineering time, and a conversation whose content the person could have had in March.

The rule the whole thing rests on

Nothing in a formal conversation should ever be new information.

If someone learns in a review that their communication has been a problem for six months, the failure isn't the review. It's the six months, and it belongs to their manager. The review just made it visible, which is the one useful thing an otherwise wasteful process does, and you can get the same result more cheaply by holding the standard directly.

So the mechanism is the one-to-one, and it has to be good. Weekly, thirty minutes, and it's theirs: their agenda, their topics, and if they have nothing, that's a signal rather than a reason to cancel. The manager brings one thing per session at most. Not a status update, which they can read. Something they noticed.

The half that most managers skip is the specific and the immediate. Feedback delivered within a couple of days of the thing, naming the thing, is worth ten times the same feedback in a quarterly summary. "In that review yesterday, the way you responded to Sam's question closed the conversation down" is usable. "You can be a bit dismissive sometimes" is not, and it will be heard as an attack on character rather than a note about a behaviour.

Twice a year, have a longer conversation that steps back: where is this going, what do you want, what's the gap between here and the next level. Not a review. No form. Ninety minutes and a shared document you both edit afterwards, so there's a written record of what was agreed, because memory of these conversations diverges dramatically within about a month.

The one formal thing I keep

Written 360 feedback, on a regular cycle, for everyone.

That's inconsistent with everything above and I know how it looks. Here's why it isn't.

Continuous one-to-one feedback has a structural blind spot: it all flows through one person's view. A manager sees a fraction of what someone actually does, and the fraction they see is biased toward the things that reach them, which is disproportionately problems and demos. What they don't see is what it's like to be in a code review with this person at 6pm, or to be the junior engineer who asked them for help twice.

Peers see that. Nobody else does. And peer information does not surface on its own, because volunteering it feels like telling on someone, so it requires a structured, scheduled invitation. That's what the 360 is for: not a rating, not an input to compensation, just a periodic collection of what the people around someone actually experience.

The rules that keep it honest. Feedback goes to the person, not just to their manager. It asks for specifics, and vague submissions get sent back. It's separated in time from any pay conversation, by at least a couple of months, because the moment it feeds a number people start managing it. And it names things to keep doing as well as things to change, which sounds like a nicety and is actually the part that makes people read the rest of it.

What it catches, reliably, is the strong performer that nobody wants to work with, and the quiet person who is holding up two teams and has never once mentioned it. Both are invisible to a manager, both are obvious to peers, and both are expensive to keep missing.

Calibration, and what it corrupts

Calibration is the meeting where managers compare their people against each other to normalise standards. It fixes a real problem: one manager's "exceeding" is another's "solid," and without a comparison the ratings mean nothing across teams.

It corrupts things in two specific ways, and both are worth knowing before you introduce it.

It rewards visibility, because a manager arguing for their person in a room of peers is doing advocacy, and the people with legible, demonstrable work are much easier to advocate for than the person who prevented three outages nobody had. Every calibration room systematically underrates the maintainers.

And it drifts toward a distribution even when nobody has mandated one. Put eight managers in a room to compare thirty people and a curve appears by social pressure alone. Stack ranking without the paperwork.

If you have more than about six managers, you probably need some version of it, because the alternative is worse. Below that, two managers and a director talking honestly about specific people, with examples, covers it. Keep it about the standard rather than about the distribution, and be willing to say out loud that this team genuinely does have four strong people.

The strong performer everyone dislikes

The case that tests whether your system is real.

Somebody delivers, visibly and consistently, and the cost of them lands on other people: reviews that make juniors afraid to submit, meetings where the discussion ends when they speak, three people who've quietly asked not to be on projects with them.

Two things are true at once. Their output is real, and you're paying for it out of an account you can't see. The people leaving because of them don't cite them in the exit conversation. They say they wanted a new challenge.

Handle it as behaviour, immediately and specifically, in the one-to-one, with the actual examples: this review, this comment, this meeting. Not "some people find you difficult," which is unanswerable and immediately becomes a hunt for who complained. And be clear that it's a condition of the job rather than a preference, because the version where it's raised gently every few months for two years is the version where the team learns that output buys immunity.

Some of them change, and quite a few do once someone finally tells them plainly. The ones who don't, you eventually have to lose, and you'll lose them later than you should have because their output makes the arithmetic feel wrong right up until you count the people who left.

If you're going to keep a formal cycle anyway

Plenty of companies can't remove it. HR requires it, or the board expects it, or the org is big enough that consistency genuinely needs a container.

Then make it cheap and make it a summary. No self-assessment essays. Manager writes half a page drawn from things already discussed, the person reads it before the meeting rather than during, and the meeting is spent on what happens next rather than on relitigating the past year.

And keep the pay conversation on a different day from the feedback conversation. Held together, the person hears the number and stops listening, and everything you said about their growth is gone.