Deals that have already resolved (acquired, IPO'd, or written down) are kept separate from active opportunities — they inform the model, they aren't something you can still invest in.
How the composite score is built, what it doesn't yet account for, and every known data-quality issue in this dataset.
These are the spec's starting priors, not yet informed by real outcomes.
Every score shows how many of the 4 weighted factors were actually available for that deal — not just whether the deal "looks" complete.
4/4 — full data: comps-derived valuation delta, founder score, investor score, and a secondary-market reference price all present.2–3/4 — partial: usually missing valuation_delta_vs_comps, because deal emails almost never disclose revenue, so there's nothing to compare against public comps.Surfaced automatically from enrichment/scoring notes across the dataset — not exhaustive, but everything flagged so far:
key_leader_score (for de facto leaders who aren't legal founders, e.g. a President/CTO who joined post-founding) is tracked but not yet wired into the composite formula.