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.
Direct answer
In the SSRN sample, a sustained commit-velocity spike (over 40% on a 14-day rolling window versus the prior 90-day baseline) precedes the public fundraise announcement by a median of 7 weeks, with a 90% confidence interval of 4-13 weeks. The threshold catches roughly 60% of subsequent $1M+ rounds at about a 22% false-positive rate.
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).
The practical value of the window is not precision but lead time. A median of 7 weeks with a 4 to 13 week interval means the signal is too coarse to time a meeting to the week, but more than good enough to convert a passive watchlist into an active outreach queue, because it consistently fires weeks before the press release that everyone else sees. The right use is to treat the alert as the start of a working window, not a countdown timer.
The threshold carries an explicit tradeoff worth internalizing. The 40 percent spike over a 14-day window against a 90-day baseline catches roughly 60 percent of subsequent $1M+ rounds at a false-positive rate around 22 percent. Tightening the threshold cuts false positives but loses recall, and loosening it does the reverse, so there is no free lunch in choosing a cutoff, only a deliberate preference between missing fewer rounds and reviewing fewer duds.
The window also works in reverse for diligence, which is the under-used direction. A startup announcing a round today that showed no commit-velocity spike in the prior 12 weeks is not necessarily a problem, private repos and pre-product rounds are legitimate, but the absence of the signal does reduce the external verification a backer can gather from public engineering output. Forward, the window is a sourcing advantage; backward, it is a consistency check.
Reproducibility is the reason to trust the number at all. The entire computation is public: pull GitHub activity, compute a 14-day rolling velocity against a 90-day baseline, threshold at plus 40 percent, and cross-reference fundraise announcements weeks later. The methodology and the SSRN preprint document it end to end, so a fund can re-run the analysis on its own pipeline rather than accept the headline figure as given.
Quote-ready takeaway
In the SSRN sample, a sustained GitHub commit-velocity spike (over 40% on a 14-day rolling window versus the prior 90-day baseline) precedes the public fundraise announcement by a median of 7 weeks, with a 90% confidence interval of 4-13 weeks. The threshold catches roughly 60% of subsequent $1M+ rounds at about a 22% false-positive rate.
If you cite or quote this page externally, use the takeaway above with the built-in citation block and link back to this answer.
Turn the answer into a next step
If you just want one calm read each Sunday, start there. If the question is already expensive, use First Look. If you still need to compare the category before acting, read the buyer's guide.
Already comparing tools? Read the buyer's guide or test one sector with First Look (€7).
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.