Performance review: making the meeting worth the preparation

A performance review is a scheduled conversation between someone and their manager about how the work has gone, what should change, and what happens next — with a written record that both sides agree reflects what was said. Almost every complaint about the format comes down to one thing: something appeared in the review that had never been said before.

That is worth stating as a rule, because everything else follows from it. A review should contain no new information. If a problem has been going on for four months, the review is far too late to raise it, and raising it there converts a routine conversation into an ambush that damages the working relationship more than the problem did.

What a performance review has to cover

• What was agreed last time, and what happened to it. Skipping this teaches everyone that the review has no consequences.
• The work itself, discussed with specific examples rather than adjectives.
• What the person does well, in enough detail to be repeatable — good on client work is not usable feedback.
• What needs to change, with what good would look like and by when.
• Support required: training, tools, a decision from someone else, a blocker removed.
• What the person wants next, which is the half most reviews leave out.
• A short written summary agreed by both, and a place for the employee to disagree on the record.

The single most useful question is the one that sounds soft: what is getting in your way? It surfaces process problems, unclear ownership and tooling gaps that a manager rarely sees from where they sit, and it is answered honestly far more often than what would you like to improve.

Ratings and what they do to the conversation

Numeric ratings make the review easier to aggregate and harder to have. The moment a score is on the table, the person listens for the number, decides whether it is fair, and stops absorbing anything else — especially if pay is attached to it. That is not irrationality. It is a sensible response to the fact that the number, not the discussion, is what travels upwards.

If ratings are required, two things reduce the damage. Separate the development conversation from the pay conversation by several weeks, so each can be had properly. And publish what each rating level means in behavioural terms, so a three is a definition rather than a verdict. Forced distributions — a fixed quota of each rating — do more damage than the ratings themselves, because they make the outcome partly a function of who else happens to be in the group.

Running the cycle

1. Keep brief notes through the year, both directions, so the review is assembled rather than remembered.
2. Send the questions a week ahead and ask for a written self-assessment first.
3. Read the self-assessment before writing your own view, not after.
4. Hold the meeting somewhere private, with the phone away and no hard stop ten minutes later.
5. Spend most of the time on what is next, not on relitigating what happened.
6. Agree at most three things to change, each with what support comes with it.
7. Write it up within two days, share the draft, and let the person add their own comments before it is filed.
8. Diarise a short check-in at the halfway point, which is what turns the agreement into something real.

Frequency

Annual reviews persist because they fit the pay cycle, not because twelve months is a useful feedback interval. A quarterly or half-yearly rhythm catches problems while they are still small, and shortens the memory required. Many small companies land on a light quarterly conversation with one longer annual one, which works as long as the light version is not skipped whenever things get busy.

Whatever the cadence, the reviews are also a record with legal weight. In most jurisdictions they are the evidence for whether a performance concern was raised, when, and what support followed — so accuracy matters, as does letting the employee record disagreement. Where you are considering action based on performance, the applicable employment rules govern the process, and that is a conversation for an adviser rather than a template.

Keeping the record

Ettex Records suits the record side of this: one entry per review, with the date, the agreed actions, the check-in and the written summary attached, so a year later the sequence is visible rather than scattered across documents and inboxes. The half-year check-in becomes a dated item rather than an intention.

It is not an HR system. No rating scales, no calibration, no anonymous feedback collection, no compensation data, and no permissions model built for employee files specifically — access is the same sharing model as anything else, which is worth knowing before you put sensitive material in it. Where reviews contain personal data, who can see them is a decision to make deliberately.

No surprises, examples rather than adjectives, most of the time spent on what is next, and a check-in in the diary before anyone leaves the room.