Identifies the top 5 most impactful engineers on posthog/posthog over the last 90 days, using GitHub collaboration data (not lines of code).
https://kazche.github.io/posthog-engineers-analysis/docs
Impact = a 0–100 score blending four percentile-ranked signals, each shown transparently on the dashboard:
| Signal | Weight | What it captures |
|---|---|---|
| Reviews given | 33% | Unblocking & mentoring (depth-weighted; rubber-stamps count less) |
| Shipped work | 28% | Merged PRs + issues closed (reach) — never LOC |
| Cross-area breadth | 22% | Distinct subsystems shipped to |
| Collaboration | 17% | Comments on others' work + fast review turnaround |
See APPROACH.md for the full methodology and caveats.
scripts/pull.py # GraphQL pull of PRs created in the 90-day window -> data-dump/raw_prs.json
scripts/analyze.py # scoring + percentile normalization -> data-dump/scores.json
scripts/build_dashboard.py # inlines scores.json into a single static page -> docs/index.html
scripts/sanity.py # validates top-5 counts vs GitHub Search API
Data: 11,050 PRs analyzed, 140 active engineers. Top-5 merged counts validated against GitHub to within 0–2 PRs.