Talent decisions need a record, not a memory

The default talent conversation in a company of fifteen people happens like this: six months of work, judged in twenty minutes, from memory. The week before the conversation shapes it more than the nine months before that.

Why memory fails

Recency is the strongest force in any review. The project that just ended, the meeting from last Tuesday, the escalation from Friday — those are what people remember. Not the decision that saved a client in March, and not the support someone quietly gave a new colleague all summer.

A second failure gets less attention: disagreement between managers is usually not about the person, it is about the role. When two managers describe the same person differently, they often hold two different ideas of what the role expects — and neither has been written down.

The cheapest piece of people infrastructure

A written record costs almost nothing and fixes both problems. One short note after a project or a one-to-one: what was agreed, what changed, what the person did. Not a diary and not a performance system — a log.

Over two or three cycles, the log becomes the most useful document in the company. It shows growth, it shows promises, and it turns the next conversation into an update instead of a debate.

Calibration, not ranking

The historical warning here is well documented. Once grading becomes a distribution — a fixed share of top ratings, a fixed share of exits — the curve starts driving decisions instead of describing them. Dick Grote's The Case Against Forced Ranking (Harvard Business Review, 2005) described the consequences in companies that tried it: collaboration drops, managers optimise for the ranking conversation, and trust goes with it.

Gallup's workplace research supports the other half of the argument: the manager relationship is where engagement is made or lost, with managers accounting for at least 70% of the variance in engagement scores across teams. A process that damages that relationship costs more than it measures.

So the rule we recommend is deliberately modest:

• Define the categories you use internally, and define which actions each one allows.
• Put evidence behind every rating, written at the time it happened.
• Calibrate differences between managers openly, in one room.
• Write the outcome down with a date, so the next audit starts from the last agreement.

That is the whole process. Everything else in this space — dashboards, competency libraries, nine-box slides — is decoration until those four things exist.

None of the four steps is hard; they are simply easier to skip than to do. How to run performance reviews in a small business, how to evaluate employee performance fairly, and how to keep performance reviews for small teams from turning into an argument about perception are really one question: was there a standard, written down, before the outcome was known?