Once a review is open, both the employee and their manager have an assessment to write. It's the same screen and the same wording on both sides — a manager doesn't get a more clinical version of the form than the person being assessed writes for themselves. Only the name on it changes.
Writing blind, on purpose
Neither side can read what the other has written until both have submitted. There's no override, at any level, for anyone. Seeing your manager's draft before finishing your own turns your assessment into a reaction to theirs instead of an independent one, and independence is the point: blind-then-reveal is what makes the differences at alignment real differences, rather than one side quietly agreeing with the other in advance.
One dimension at a time
The form walks through each dimension in the current framework, showing the expectation text for the person's level right beside where you're rating it, not on a separate page you'd have to hold in your head. For each one, pick a rating and leave a note — a sentence on what you'd say out loud, ideally naming something specific that happened and what came of it, rather than a general impression.
The rating scale
Five ratings, worded around what's been observed rather than the person's worth:
- Developing — not meeting the expectation consistently yet; needs support, practice, or a different approach.
- Consistent — meets the expectation reliably, without prompting.
- Strength — clearly above the bar here, and others benefit from it.
- Next level — repeatedly does what the next level up expects, at the right scope. (This doesn't appear at the top of a track — there's nowhere higher to reach.)
- Not observed — there was no fair chance to see this one.
Clicking a rating you've already chosen clears it. "Not rated" is a real state here, not something the product assumes for you by default.
Evidence
Ratings of Developing, Strength, or Next level, anything off the middle of the scale, require at least one piece of evidence before you can submit. Consistent and Not observed don't, because the claim behind them is smaller. Write a fresh entry on the spot, or cite something already logged, including something a teammate logged and shared with you. See Evidence for how logging works, and what happens if you try to cite something that hasn't been shared.
This is what keeps "no evidence" and "weak performance" from ever looking the same on the page: an unsupported claim doesn't get to sit at the same weight as one backed by something that happened.
Submitting
Your draft autosaves as you go. There's no manual save button, just a status line showing it's saved, or retrying if a save failed. Submit is the one deliberate action: your side becomes visible to the other person the moment they've also submitted, and it's final. There's no unsubmitting.
If the workspace's subscription has lapsed, the assessment screen goes read-only, but your draft is kept exactly as it was rather than lost. See Settings & billing for what read-only means workspace-wide.