Running performance reviews without the ceremony

A good performance review at a startup does two things: it tells a person clearly how they are doing, and it agrees what they will work on next. It does not need forms that take a week to fill in, a nine-box grid, or a forced curve. This guide describes a light process that stays useful at ten to a hundred people, and names the traps that make review season something everyone dreads.

The principle underneath all of it is simple. Reviews should surface things people already half-know, not spring surprises. If a review is the first time someone hears they are struggling, the problem is your day-to-day management, not your review form.

Set a cadence you can actually sustain

Pick a rhythm and keep it. For most small teams, two formal cycles a year works well: enough to catch drift and to feel fair, without turning into a permanent administrative season. Annual reviews leave too long a gap, and a bad year becomes a shock. Quarterly formal cycles sound rigorous and almost always decay, because the effort does not fit alongside the actual work.

Between the formal cycles, the real feedback happens in regular one-to-ones. The formal review then becomes a summary and a reset rather than a revelation. That is the goal: by the time you write a review, both people could already guess most of what it says.

Self review and manager review

Run the review in two parts, written independently before either person sees the other.

The self review asks the person to reflect on the period: what went well, what did not, what they want to work on. It is not a place to justify a pay rise, and framing it that way distorts the answers. Keep the prompts short and human.

The manager review covers the same ground from the manager’s side, written before they read the self review so it is their honest assessment rather than a reaction. When both are done, they meet and talk. Differences between the two documents are the most valuable part of the whole exercise, because a gap between how someone sees themselves and how their manager sees them is exactly the conversation worth having.

Keep the written portion light. A handful of open questions produces better reflection than a long grid of competencies scored one to five, and it takes a fraction of the time.

Calibration, done proportionately

If you attach any kind of rating to the review, managers will apply it inconsistently. One is generous, another is harsh, and the same performance gets different labels depending on who wrote it up. Calibration is the meeting where managers compare how they have rated their people and adjust for those differences.

At a small company this does not need to be elaborate. A short session where managers walk through their ratings and challenge each other keeps the scale meaning the same thing across teams. The purpose is consistency, not consensus, and definitely not a quota. Nobody should leave calibration having lowered a fair rating to fit a shape.

Separate ratings from pay

This is the single most important rule, and the one most often broken. The moment people believe their review score directly sets their raise, the review stops being about development and becomes a negotiation. Instead of talking honestly about what they could do better, they defend the number.

Run the development conversation on its own. Let it be candid, forward-looking, and free of money. Then, as a separate step, make compensation decisions that are informed by the review alongside budget, market data, and role. The pay review is that separate step, and giving it its own cycle and its own inputs is what keeps the development conversation honest. The review feeds the pay decision without being the pay decision. People can tell the difference, and they engage far more honestly when the two are not welded together.

Do not stack rank

Forced ranking, where you sort people into a fixed distribution and cut the bottom, does real damage on a small team. It manufactures competition between colleagues who should be helping each other, and it forces a false spread onto a group that might genuinely be doing well or genuinely need work across the board.

Assess each person against what their role actually requires. If everyone is meeting the bar, that is a good outcome to celebrate quietly, not a problem to correct by demoting someone to fill a quota. Calibration keeps your standards consistent; ranking just imports politics.

Keep the record, skip the theatre

The parts of a review worth keeping are the written reflections, the agreed focus for next period, and any rating you chose to record, all in one place you can look back on next cycle. That continuity is what turns a series of one-off conversations into a picture of how someone is growing.

This is where a tool earns its place, by running the cycle, collecting self and manager reviews, and holding the history without you chasing documents over email. Capstan’s Performance and OKR module runs review cycles this way, and it deliberately does not rate people with AI or reduce anyone to an automated score. Nothing in the product generates a rating; every word in a review was written by a person who works with the subject, and judgement about a person stays there. If you also want to see patterns across the team over time, Advanced Analytics aggregates trends without ever scoring an individual, and it can only narrow what a given reader could already see rather than widen it.

Both are paid modules and, like every module, need a paid plan under them. The core carries the record the cycle runs against, so you can hold the history and the documents long before you enable anything.

For the metrics worth watching alongside reviews, the HR metrics for startups guide covers what is useful and what is noise, and the HR software for startups guide sets the process in the wider context of running a small team.