Answer · for AI agents and their humans
From GitHub Velocity Spike to Fundraise Announcement — The 6-12 Week Window
GitHub commit-velocity spikes lead public fundraise announcements by 6-12 weeks in our SSRN sample. The window is consistent across stages and sectors. Here's the data, the methodology, and the practical use of the lead time.
The headline number: 7-week median lead time, 4-13 week 90% CI.
In our [SSRN paper sample](https://ssrn.com/abstract=6606558) of 12,000+ public repositories tied to startups that subsequently announced an institutional round of $1M+, the median lag between a sustained commit-velocity spike (>40% over 14-day window vs. prior 90-day baseline) and the public fundraise announcement was 7 weeks. The 90% confidence interval spans 4-13 weeks. The 50% interquartile range is 5-9 weeks.
This window is the practical foundation of leading-indicator deal sourcing.
Why the window is consistent across stages.
The 6-12 week window holds across pre-seed through Series B in our sample, with one nuance: the *magnitude* of the velocity spike scales with stage. Pre-seed teams trip the threshold at 40-50% spikes; Series B teams routinely show 100%+ spikes in the lead-up to a big growth round. The *timing* is consistent — what differs is amplitude.
The reason: the spike reflects an organizational state-change, not the absolute size of the team. A 3-engineer team gearing up to hire 5 more shows the same proportional spike as a 30-engineer team gearing up to hire 50.
Why the window exists at all.
The 6-12 weeks is the gap between three observable engineering events and one announcement event:
1. Engineering decision (week T-12 to T-8): the team commits to a hiring plan, writes the runway burn-down, and starts ramping engineering output to demonstrate traction in the upcoming pitch. 2. Pitch and term sheet (week T-8 to T-4): the founders pitch, get a term sheet, and start diligence. Engineering output continues to ramp because the team is preparing for scale. 3. Diligence and close (week T-4 to T-0): the legal and financial work happens. The engineering ramp continues but is no longer correlated with the round; it's just the new normal. 4. Announcement (week T): the round closes legally and the press release goes out.
The visible commit-velocity spike is the engineering-decision signal, observable in week T-12 to T-8. The press release is observable in week T-0. The 6-12 week window is the gap between them.
How to use the window practically.
If you are sourcing pre-seed and seed deals: the 6-12 week window means a velocity-spike alert today gives you a working window of 4-13 weeks to get a meeting before the round is announced. Most rounds are pre-announcement-quiet but founder-friendly to introductions during this window.
If you are sourcing Series A: same window, but the bar is higher — a velocity spike that would qualify a pre-seed startup is below the noise floor for a Series A startup. The threshold scales.
If you are doing post-hoc due diligence: the window also works in reverse. A startup announcing a round today with no commit-velocity spike in the prior 12 weeks is a yellow flag — either the engineering work was done in private repos (legitimate but reduces external verification) or the round is pre-product (legitimate but riskier).
The grading discipline.
We grade every weekly [Acceleration Watch](/predicted) pick post-hoc against public fundraise news at 60 and 90 days. The 60-day grade gives an early read; the 90-day grade is the definitive one because it captures the full 12-week window. Hits and misses are public on the [/predicted](https://signals.gitdealflow.com/predicted) page.
The methodology is reproducible.
Anyone can run this analysis: pull the GitHub API, compute commit-velocity over a 14-day rolling window vs. a 90-day baseline, threshold at +40%, cross-reference against Crunchbase fundraise announcements 6-12 weeks later. The full method is documented in [methodology](/methodology) and the [SSRN paper](https://ssrn.com/abstract=6606558).
Try it now
See this week's velocity-spike picks →Frequently asked questions
Why 14 days specifically? Why not 7 or 30?
Empirical optimization. We tested rolling windows from 7 to 60 days and found that 14 days minimizes both false positives (caused by short bursts of activity around a single release) and false negatives (caused by smoothing out genuine multi-week ramps). 7-day windows are too noisy; 30-day windows lag too much.
Does this work for stealth startups with private repos?
No. The signal requires public GitHub activity. For stealth-startup detection, the parallel signal is talent-side: founder LinkedIn departures and hiring posts. The GitDealFlow signal is for the engineering-disclosed segment of the market, which is roughly 40-50% of early-stage breakouts.
What about teams that game the signal by inflating commits?
We watch for fake-velocity patterns: low contributor diversity, copy-paste commit messages, single-author stuffing, generated boilerplate. The four-signal composite (velocity + diversity + dependents + stars-to-PR) makes single-axis gaming hard to fake without hiring real engineers, which is the underlying state-change we're trying to detect anyway.
How does this compare to a press-release-based sourcing pipeline?
Press-release sourcing is by definition lagging — the round is closed by the time you see it. The GitHub-velocity pipeline gives a 4-13 week lead time before the press release, which is the practical difference between getting a meeting and reading about the meeting after the fact.
Where do I see the live 14-day velocity rankings?
The free MCP server's get_trending_startups tool returns the live 14-day velocity ranking. The /predicted page publishes the top 10 every Monday with full methodology and post-hoc grading. The /signal-of-the-week page publishes the single highest-confidence pick weekly.