If You Are Going to Do Performance Reviews, Do These Three Things

Performance review processes consume enormous organizational effort and produce, in most companies, modest value: ratings people dispute, documentation nobody reads, and conversations both parties dread. Abolishing them entirely has its own problems, chiefly that pay and promotion decisions then happen with no visible basis at all.

Three changes account for most of the improvement available.

One: separate the development conversation from the compensation decision

When both happen in the same meeting, the development conversation is worthless. Nobody discusses their weaknesses honestly while a number is being determined in the same hour.

Hold the pay and rating conversation as its own short, direct meeting: here is the decision, here is the reasoning. Hold the growth conversation at a different time of year, with no stakes attached, and make it genuinely two-way.

Two: write the evidence throughout the year, not in November

Reviews written from memory are dominated by recency and by whatever is easy to recall, which systematically favors visible work over important work and confident communicators over quiet contributors.

A running file per person — a few lines a month on what they did, what landed, what did not — takes a manager under two hours a year and transforms both the accuracy and the credibility of the outcome. It also makes the conversation far easier, because it is grounded in specifics rather than impressions.

If a manager needs three days to write a review, the problem is not the form. It is that they were not paying attention until October.

Three: eliminate surprise

Any rating that surprises the recipient represents a management failure earlier in the year. The review should be a summary of conversations already held, not the first delivery of significant information.

A practical rule: nothing appears in a written review that has not been said to the person's face at least once before. Managers who adopt this find their difficult reviews get easier, because the hard part happened in March when it could still be addressed.

What to cut

• Numeric precision beyond what you can defend. Three or four levels are enough; nine create arguments about distinctions nobody can articulate.
• Self-assessments that simply become the manager's draft. Ask instead for two accomplishments and one thing the person wants to develop.
• Long-form peer feedback collected annually. It is inflated, time-consuming, and less useful than two targeted questions asked of two colleagues.
• Forced distributions applied to small teams, where they produce visible arbitrariness.

What to keep

A written record of expectations and outcomes, a clear rating with a stated basis, a documented development priority, and the person's own comments. That is a defensible process, and it can be delivered in a two-page form.

The purpose of the exercise is to make judgment explicit and reviewable. Everything in the process that does not serve that purpose is administrative theater, and cutting it is the fastest way to get managers to take the remainder seriously.