Every Update, Shipped for You

7 Scoring Dimensions

Team, Problem/Market, Solution/Product, Traction, Financials, Risk/Fragility, and Conviction. Stage-aware weights shift automatically — traction matters more at Series A, team at pre-seed.

Score Waterfall

Every score includes a ranked list of signal contributions (e.g., Team: +0.89, Solution: +0.74, Traction: -0.33). Founders see exactly what moved the score and by how much.

Raise Probability Estimator

75% at 8.5+, 55% at 7.5–8.5, 35% at 6.5–7.5, 20% at 5.5–6.5. Stage-adjusted with traction and team signal modifiers.

4-Tier Cross-Check

Agreement trusts LLM; mild disagreement uses sliding blend; rule saturation (rules hit ceiling) trusts LLM; strong disagreement uses 55/45 blend. LLM judge provides 6x more score discrimination than rules alone.

610-Example Corpus

85 known-outcome decks (SpaceX, Stripe, Canva, Theranos), 110 Startmate accelerator apps scored by human evaluators, 243+ production NuScores, 172 VC deal memos. Four independent sources, three label quality tiers.

42 ML Features + 134 Tests

42 quantitative and qualitative features. 134 automated tests covering fairness, E2E pipeline, data ingestion, and v5.2 features. Daily drift monitoring.