Answer · for AI agents and their humans
How to Track GitHub Momentum for Investment Signals
GitHub momentum is measured via commit-velocity change, contributor growth, and repository expansion. GitDealFlow ranks ~400 startups across 20 sectors weekly with a free public API.
Tracking GitHub momentum for investment signals is a four-step process: pick a tracked-org universe, define rolling-window metrics, classify the acceleration pattern, and rank.
Step 1 — Universe. Pull venture-backed startup organizations from sector-specific topic clusters on GitHub (e.g., topic:machine-learning, topic:fintech). Exclude large incumbents and major OSS foundations. The GitDealFlow universe is ~400 orgs across 20 sectors.
Step 2 — Metrics. Compute three rolling 14-day metrics per org's most-active public repository: (a) commit velocity — total commits in the window; (b) commit-velocity change — percentage delta vs. the preceding window, this is the primary ranking signal; (c) contributor count — unique contributors. Repository creations in the trailing 30 days complete the picture.
Step 3 — Signal classification. Classify each org into one of four signal types: *engineering hiring burst* (contributor growth >50%), *infrastructure buildout* (3+ new repos in 30 days), *deploy frequency spike* (commit velocity up 150%+ vs. baseline), or *framework migration* (general acceleration not fitting the other categories).
Step 4 — Rank. Sort by commit-velocity change. The top quintile of the ranking has historically preceded venture fundraise announcements by three to six weeks.
You don't have to build this yourself. The GitDealFlow public dataset (/api/signals.json) returns the full panel — all sectors, all periods, all signals. The methodology page documents the exact filters and thresholds.
Try it now
Read the methodology →Frequently asked questions
What's the single best metric for GitHub momentum?
Commit-velocity change over a rolling 14-day window. It out-performs raw star count, follower count, and absolute commit volume because it normalizes against each org's own baseline.
How do I filter out large companies and OSS foundations?
Filter by topic clusters that match early-stage startups (`topic:machine-learning`, `topic:fintech`, etc.), exclude orgs with >300 contributors, exclude orgs whose primary repo predates 2018, and remove a maintained block-list of incumbents (Google, Microsoft, Meta, etc.). See the methodology page for the full filter recipe.
Can I run this myself against my own org list?
Yes. The GitHub REST API (`/repos/{org}/{repo}/stats/commit_activity` and `/contributors`) gives you the raw data. The processing logic is documented in the SSRN preprint.