Performance reviews that people do not dread

Most performance reviews fail for one of two reasons: nobody wrote anything down during the year, so the review is a recency-biased guess, or the review and the increment are the same conversation, so nobody tells the truth. Both are fixable without a large HR team. This guide lays out a cycle that a 20 to 200 person company can run twice a year with one spreadsheet and a calendar.

Decide what the review is for

Pick one primary purpose and say it out loud. The usual candidates are development (what should this person work on next), decisions (confirmation, promotion, increment) and documentation (a record of how someone performed). A single form cannot do all three well, because the moment a rating affects pay, the self-review becomes advocacy.

The pragmatic split that works for small companies: a half-yearly development review in September or October with no money attached, and an annual review in March that feeds the April increment. The first builds the habit of honest conversation; the second uses the record that habit created.

The cycle, step by step

1. Goals set in April, written in two to five lines per person. Not a cascade of OKRs. Three outcomes the person and their manager agree matter this year, plus one development area. Store them where both can see them all year.
2. Notes during the year. The manager keeps a dated, two-line note after every meaningful win, miss or conversation. Ten minutes a month per report. This is the whole difference between a fair review and an unfair one; everything else is formatting.
3. Self-review, one week, five questions. What did you deliver against your goals. What are you proudest of. What did not go well and why. What do you want to do more of. What do you need from your manager. Cap it at a page; a long self-review rewards writing ability, not performance.
4. Manager review, one week. The manager writes against the same goals, citing the notes. Where the manager and the employee disagree about a fact, resolve the fact before the meeting, not during it.
5. Calibration, one meeting. Managers of similar teams compare their ratings before anything is shared with employees. This is where a lenient manager and a harsh one are brought to the same scale. In a small company it is a single two-hour meeting with the founders present.
6. The conversation, 45 minutes. The written review is shared a day in advance so nobody is reading it live. The meeting is for discussion, not announcement.
7. Sign-off and storage. Both sign, the record goes into the employee's file, and the next cycle's goals are written in the same sitting while context is fresh. If your HR system keeps an employee timeline, attach it there so the next manager can find it.

Ratings: fewer, defined, and anchored

A five-point scale is standard, but the labels matter more than the number of points. Define each level with a sentence that describes behaviour, not adjectives. "Consistently delivers goals with little supervision and helps others deliver theirs" is a definition; "Exceeds expectations" is a feeling.

Two rules that prevent most arguments. First, a rating must be supported by at least one written example from the year's notes. Second, the top and bottom ratings require a second manager's agreement during calibration. Forced distributions (a fixed percentage must be rated low) are a bad idea below a few hundred people; the statistics do not work on a team of six, and everyone knows it.

Separating the review from the money

The annual review informs the increment, but the number should come from a pay framework, not from the conversation. Decide the increment budget first, map rating bands to increment ranges, and apply the ranges after calibration. The manager then delivers two pieces of information at two different times: the review in March, the increment letter in April. If the review is honest and the increment is consistent with it, the second conversation is short. See our guide to pay bands and increments for the framework side.

Common failure modes

• No notes. The review becomes about the last six weeks. Fix by making the monthly note a manager habit, checked in 1:1s with their own manager.
• Surprise feedback. Nothing in a review should be heard for the first time in the review. If it is, the manager failed earlier in the year, and the review is not the place to make up for it.
• Reviews that never close. Set hard dates and publish them. A cycle that drifts into May undermines the increment date and signals that it does not matter.
• One manager, no calibration. A single person's scale drifts over years. Even in a company with two managers, compare before publishing.
• Attrition in the gap. Resignations spike after increments are announced, not before. Honest reviews reduce the shock; they do not remove it.

A good review cycle is boring: the same dates, the same five questions, a record that grows every year. The companies that get it right are rarely the ones with the most elaborate form. They are the ones where a manager can open last year's file and show the employee exactly why this year's rating is what it is.