Week 4: The self-review as fiction

The brag document, the self-review, and Amazon's 2026 requirement that every corporate employee list three to five accomplishments with impact.

16 March 2027

Why it worksSelf-ratings run half a standard deviation above the manager's, and the summary that reaches the room keeps the numbers and drops the sentences.

Julia Evans’s advice is to keep a running document of what you did, so that when the review comes your manager can make your case from it and a new manager can learn what you did before they arrived. It is good advice. It is also the raw material of the self-review, which is the one document you write that reaches the calibration room.

The document has a known bias, and it was measured before most of you were born. Harris and Schaubroeck’s 1988 meta-analysis put self-ratings more than half a standard deviation above the ratings supervisors gave the same people, with a correlation between the two of about a third. Everyone who writes a self-review rates themselves generously, so the generous version is the baseline, and the honest version reads as a confession. That is the sense in which the self-review is fiction: not that anyone lies, but that the genre has a house style and the house style is inflation.

In January 2026 Amazon’s review process began requiring corporate employees to list three to five accomplishments demonstrating impact, feeding directly into compensation. That is the self-review with the pretence removed: a short list, summarised again by the manager, summarised again by the chair. Each summary keeps what is easy to carry, and what is easy to carry is a number. This lecture is about what that does to the truth of the document, and why the honest version so often loses to the quantified one.

Outline#

• the brag document, and why it works even when nobody games it
• half a standard deviation: what the research says every self-rating does
• Amazon’s three to five accomplishments, and what “impact” means on a form
• three summaries deep: what survives and what does not
• the line between a defensible estimate and a made-up number

Sources

• Harris and Schaubroeck, A meta-analysis of self-supervisor, self-peer, and peer-supervisor ratings (Personnel Psychology, 1988)
• Julia Evans, Get your work recognized: write a brag document
• Amazon asks employees for a list of accomplishments (Fortune, January 2026)