How impact is scored
Not lines of code or raw counts. Each 0–100 score blends four signals, each ranked as a
percentile vs. all active engineers — so every point is traceable.
Reviews given 33% — unblocking & mentoring.
Reviews left on teammates' PRs. A review that requests changes or leaves real comments counts
more than a bare “approve” (a rubber-stamp), so thoughtful reviewers rank higher.
Shipped work 28% — delivery.
Pull requests merged, with extra credit for ones that closed a tracked issue (real reach,
not just churn). We never count lines of code.
Cross-area breadth 22% — versatility.
How many distinct parts of the codebase they shipped to (frontend, backend, plugin-server,
Rust, products, …). Rewards system-wide impact over working a single corner.
Collaboration 17% — helpfulness & speed.
Comments on others' work, plus how fast they turn reviews around (within 24h of a PR being ready).