A Deal Flow Scoring Framework: Rank Inbound Startups Without a Full Partner Meeting
A four-factor deal flow scoring framework for early-stage investors, with a fifth engineering-velocity factor from public GitHub data, plus calibration guidance.
Key Takeaway
Scoring exists to make the yes and no decisions explicit and comparable before you spend a partner meeting. This post lays out a four-factor scorecard with weights, adds engineering velocity as a fifth dimension from public GitHub data, and explains how to calibrate the numbers so the score means something.
A scorecard is not a substitute for judgment. It is a way to make judgment explicit, comparable, and reviewable, so that the decision you make on deal forty is as disciplined as the one you made on deal one.
Here is a framework that is fast enough to use on every inbound deal and structured enough to calibrate.
Score Before You Meet#
The point of scoring is to rank before you spend the expensive meeting. Score on what is public, then decide who earns the call. A 0 to 10 scale on each factor keeps the pass fast enough that you will actually do it.
The Four Core Factors#
**Team.** Do the founders have the right capability and track record for this specific problem. At seed, this is the highest-weight factor.
**Market.** Is the market real, large, and growing. A great team in a bad market is still a bad investment.
**Product.** Does the product exist and solve the stated problem. A live product beats a deck, every time.
**Traction.** Is there evidence of real usage, revenue, or engagement. Traction weight rises with stage.
The Fifth Factor: Engineering Velocity#
Public GitHub activity adds a leading-indicator dimension the other four do not cover. It reads whether the team is shipping faster or slower, independently of what the deck claims. The backtest against 219 fundraises found a 3.4x lift in a composite commit-velocity and contributor signal preceding Series A, with a 21 to 47 day lead [1].
Score it on trajectory: accelerating high, flat middle, decelerating low. It is the only factor on the card that predicts rather than reports.
Setting Weights#
Write the weights down. At pre-seed, team might be 40 percent with velocity and product splitting the rest. At Series A, traction and market rise. The weights are a policy; the discipline of writing them is what makes the scorecard useful later.
Calibrate or Discard#
A scorecard that is never checked against outcomes is superstition. Go back after six months and test whether high scores predicted raises and performance. Adjust the weights where the card was wrong.
The goal is not a perfect number. It is a decision process you can defend, review, and improve [2][3].