The Self-Review Builder: Turn Rough Accomplishment Notes Into Polished Performance Review Bullets
Why this prompt matters
Self-reviews that use vague, passive language (assisted with, was involved in, helped) get rated lower by managers even when the underlying work was significant — reviewers pattern-match specific, quantified language to impact and vague language to a passive contributor, regardless of what actually happened. A weak self-review doesn't just risk a lower rating; it becomes the written record a manager cites months later during promotion or layoff conversations, when nobody remembers the actual work — only what's on paper.
What we use it for
It's the week before performance review season, and you have a scrappy list of six months of Slack standups and sprint notes but no idea how to turn “helped ship the new onboarding flow” into something that reads well in a formal review.
Prompt
Act as a performance review coach who specializes in the XYZ formula (Accomplished [X] as measured by [Y] by doing [Z]) for turning vague work notes into polished, quantified self-review bullets. Context: My role is [YOUR JOB TITLE]. The review period is [REVIEW PERIOD, e.g., "H2 2026"]. Here are my rough, unpolished notes on what I did, in whatever form I have them (Slack standups, sprint notes, a messy list): [PASTE YOUR ROUGH NOTES HERE]. Task: For each rough note, produce two candidate bullets: (1) a "Conservative" version using only facts and numbers actually stated in my notes, and (2) a "Stretch" version showing what the bullet could become if I can supply one specific missing metric. Group the resulting bullets into three categories if the notes support it: Impact, Ownership, and Collaboration. Constraints: - Never invent a number, percentage, or outcome that isn't in my notes. If a bullet would benefit from a metric I haven't provided, write "MISSING METRIC: [specific question to answer]" instead of guessing. - Ban passive or vague verbs: helped, worked on, was involved in, assisted with, participated in. Require active ownership verbs: led, built, reduced, drove, designed, shipped, resolved. - Each bullet must be one line — if it needs two lines, it's actually two accomplishments and should be split. - Do not pad weak notes with unearned adjectives (Blizzard, seamless, robust) to compensate for a missing metric — flag the gap instead of disguising it. Output Format: For each original note, show: the raw note as I wrote it, then the Conservative bullet, then the Stretch bullet (or a MISSING METRIC flag if no stretch is honestly possible), grouped under Impact / Ownership / Collaboration headers.
Result
Role: Senior Software Engineer. Review period: H2 2026.
Impact
Raw note: "fixed a bunch of bugs in the checkout flow, customers were complaining"
- Conservative: Resolved recurring checkout-flow defects that were generating customer complaints, stabilizing the purchase path.
- Stretch: MISSING METRIC — ask support/product for the actual complaint-ticket count before and after your fixes, or the checkout error rate from monitoring. A bullet like "Reduced checkout errors by 41%, cutting related support tickets from 60/week to 12/week" is far stronger than "fixed a bunch of bugs" — go find that number before you write your review.
Ownership
Raw note: "helped with the migration to the new database"
- Conservative: MISSING METRIC — "helped with" is too vague to convert into any credible bullet, even a conservative one. Before this can become a review bullet, answer: what specifically did you own in this migration (a service, a data model, the rollback plan)? "Helped" reviewers cannot tell if you did 5% or 50% of the work.
- Stretch: Once you have that answer — for example, if you owned the write-path migration for the orders service — try: "Led the orders-service write-path migration to the new database, completing it with zero downtime across 3 weeks of phased rollout."
Collaboration
Raw note: "mentored the new hire, they're doing well now"
- Conservative: Mentored a new team member through onboarding, supporting their ramp to independent contribution.
- Stretch: MISSING METRIC — "doing well" needs a concrete marker. Did they ship their first PR in week 2 instead of the usual week 4? Did they pass their 90-day review with a specific rating? Try: "Mentored a new hire through onboarding; they shipped their first production PR in week 2, versus the team's typical 4-week ramp."
What this reveals: two of your three raw notes are currently unreviewable as written — not because the work wasn't valuable, but because the notes don't contain a number. That's exactly the gap to close before review season, not during it.
Most self-reviews fail for a specific, fixable reason: they describe activity instead of impact. "Helped with the migration" and "fixed a bunch of bugs" are both true statements about real work, but neither survives contact with a reviewer trained to look for the XYZ formula — accomplished X, as measured by Y, by doing Z. The problem usually isn't that the work lacked impact. It's that the person writing the review doesn't have the number in front of them when they sit down to write.
This prompt is built to separate two different problems that get conflated when someone tries to write a self-review in one pass: turning a vague sentence into a well-structured one, and finding a missing metric. Those require different actions — rewriting versus research — and doing both at once is why self-reviews take so long and still come out weak. The prompt forces a split: a "Conservative" bullet uses only what you already know, so you always leave with something usable even if you never go find the missing number. A "Stretch" bullet shows what the same accomplishment looks like with one specific metric attached, and names exactly what that metric is and where you'd likely find it.
The explicit ban on invented numbers is the most important constraint in the prompt. It would be trivial for an AI to write a confident-sounding bullet with a plausible-looking percentage attached — that's precisely the failure mode to avoid, since a fabricated metric in a formal review document is a credibility risk, not a shortcut. Flagging "MISSING METRIC" instead forces the actual useful output: a specific, answerable question you can take to a teammate, a dashboard, or a ticket system before the review is due.
The three-category structure — Impact, Ownership, Collaboration — exists because reviewers read for different signals in each. A bullet that's all impact with no ownership language reads like you rode someone else's work; a bullet that's all ownership with no measurable outcome reads like busywork. Sorting your raw notes into these buckets before polishing them surfaces gaps in the story you're telling about your own six months, not just gaps in grammar.