Loading...
The six atomic signal primitives. Each one is a single, well-defined measurement on public GitHub data with a stated formula, a stated decision rule, and the most common way it can mislead someone who skips the methodology.
Commits / 14d
Window: 14 days, rolling
A raw measure of engineering output. Useful as a baseline; not directly predictive on its own. The signal investors should watch is the rate-of-change (commit velocity change), not the absolute number.
Δ Velocity (% QoQ)
Window: Two adjacent 14-day windows
The primary ranking signal at VC Deal Flow Signal. Sustained acceleration historically precedes fundraise announcements by 3–6 weeks. Captures rate-of-change, so it works equally well across small and large repos.
Δ Unique Contributors
Window: Two adjacent 6-week windows
A proxy for headcount growth. Rising contributor count typically means new hires, contractors, or community contributors — usually the first observable signal that a recent fundraise is being deployed.
Hiring Burst
Window: 6-week window
Strong indicator that a startup recently closed a round and is deploying capital into engineering. For investors: probably too late for the current round, well-positioned for the next one.
Framework Swap
Window: Per commit
The dominant signal type observed: 75% of VC-backed startup GitHub signals are framework migrations, not new features. Counter-narrative to 'engineering velocity = hiring' — the dominant pattern is rewrites.
Infra Buildout
Window: 30-day window
Indicates a startup is moving from prototype to production. Often appears 4–8 weeks before a public launch announcement.
Each primitive feeds at least one of the workflows below. Decision rules, attribution chains, and reproducibility steps are documented in the linked pages.
See also: Knowledge graph · Glossary · Methodology · Citation guide · Corrections