A Rating Is a Decision

Make sure it's yours

In this issue:

→ Why a performance rating is the most consequential decision most managers make all year

→ What changes when AI helps write the review

→ Three questions to run before you hit submit

---

Most managers think of a performance rating as a summary. It isn’t. It’s a decision.

It decides who gets the stretch assignment, who gets promoted, and who quietly stops being considered. It follows a person for a year, sometimes longer.

And this year, a lot of those decisions will be drafted by a tool.

That isn’t the problem on its own. AI is good at organizing a year of notes and spotting a pattern you missed. The problem is what happens to the decision when the draft arrives finished.

The subtle shift

A blank review form forces a judgment. A drafted one invites an edit. You read it, it sounds right, you adjust a word, you submit:

→ The language is balanced

→ The examples are there

→ The rating matches the words around it

Nothing in that process asks whether you’d have landed on the same rating on your own.

The hidden risk

Ratings were never as objective as they looked. In a large study published in the Journal of Applied Psychology, more than half of the variation in performance ratings traced back to the person doing the rating, not the person being rated. The rater’s quirks outweighed the actual performance.

That’s the uncomfortable baseline. A drafted review doesn’t fix it. It hides it behind smoother language. The bias is still yours. Now it just reads better.

The leadership shift

Let the tool organize the year. Keep the verdict for yourself. Before any rating leaves your hands, you should be able to say why in one sentence of your own, with the draft closed.

If you can, the decision is yours. If you can’t, it isn’t finished.

---

THE RATING DEFENSE

Run this on every rating before you submit it. Three questions, about five minutes a person. Do it before calibration, not after.

QUESTION 1: The closed-draft test

Close the draft. In one sentence, say why this person earned this rating.

If the sentence won’t come without looking, the rating isn’t yours yet. It’s the draft’s.

QUESTION 2: The evidence test

Name one thing this person did this year that you saw or heard yourself. Not a metric from the system. Not a line the draft wrote. Something you’d describe the same way to their face.

If everything you can name came from the draft or the dashboard, you’re rating the record, not the person.

QUESTION 3: The swap test

Would you give the same rating if the work had looked exactly the same, but the person were quieter, newer, or less like you?

This is the one most people want to skip. It’s the one that catches the rater in the rating.

Reading the result

All three held. The rating is a decision you made. Submit it.

Question 1 failed. Rewrite the reason before you touch the rating. Often the rating survives and the review gets better.

Question 2 failed. Go get the evidence. One conversation this week is worth more than another paragraph.

Question 3 failed. Don’t change the rating yet. Ask someone you trust to look at it with you. That’s what calibration is supposed to be for.

Two more this week

→ In your next 1:1, ask the person what they’d call their biggest contribution this year. Compare it with what you wrote.

→ Read one review you’ve written out loud. Mark every sentence you wouldn’t say to their face.

---

Question of the week: Could I explain every rating I wrote this year with the draft closed?

---

If this landed, send it to one manager who’s writing reviews this week. Before calibration, not after.

→ Restack it so other readers here find it

→ Forward it to the manager who needs it most

Lead what AI can’t.
JoshHMiller.com