How it works
Score the proof, not the pitch.
Any deck can tell a good story — and now any founder can generate a polished one in minutes. NUVC scores against verified evidence, not the claims on the slide. Here's how a NuScore is built, and why you can trust the number.
The short answer
Does NUVC just use an AI to rephrase my deck?
No. The deck is the input, not the verdict. A NuScore fuses three independent signals — a deterministic rule engine, machine-learning features, and an AI scoring judge — and then cross-checks the claims against public data. A single language model never decides your score on its own. That's the difference between a tool that summarises what you wrote and one that evaluates whether it holds up.
Why you can trust it
How does NUVC know a claim is true?
Three things separate a NuScore from a confident guess: it verifies claims, it weights verified facts above stated ones, and it's calibrated against companies whose outcomes are already known.
Pillar 1 — Verification
Claims are checked, not taken on faith
An integrity layer runs on every deck: AI-generated-content detection, financial-consistency checks, and cross-signal contradiction analysis. If the traction slide and the financials disagree, the score knows.
Pillar 2 — Provenance
Verified facts outweigh stated ones
A fact confirmed against public sources counts for more than the same fact asserted on a slide. Enrichment runs against public data, and anything that can't be corroborated is marked unverified — never asserted as true.
Pillar 3 — Calibration
Tuned on real outcomes, never false precision
The score is calibrated against 1,400+ real funding decisions — including 85 known-outcome companies from SpaceX to Theranos. Every score ships with a confidence level. Sparse data means low confidence, and we say so.
The pipeline
How does NUVC score a pitch deck?
When you upload a deck, you trigger a workforce — eight named agents run in parallel and return a result in about 60 seconds, across 6 venture dimensions.
Extract the deck
Parse the PDF into structured signals — team, market, traction, financials, the ask.
Score with three independent signals
An AI judge, a deterministic rule engine, and machine-learning features each weigh in. Fusion reconciles them — no single model decides.
Check the claims
Integrity flags contradictions and AI-generated content; enrichment verifies team and traction against public sources, asynchronously.
Explain and calibrate
Return a 0–10 score with a confidence level and a waterfall showing which dimension moved the number.
Want the underlying science — datasets, papers, and the fund-scoring benchmark? Read the research →
The dimensions
What does a NuScore measure?
Six weighted venture dimensions, each scored 0–10 with reasoning, plus a conviction gate, combined into one fundability score:
Responsible AI
Is the scoring fair to first-time founders?
NUVC deliberately does not collect founder gender, ethnicity, age, or educational prestige — the model cannot discriminate on data it does not hold. A first-time founder in regional Australia is scored on the same merits as a repeat founder in San Francisco. Every score is explainable, every founder can appeal and request human review, and a governance layer monitors outcomes for anomalous score clustering and recalibrates when patterns disadvantage any cohort.
See what the proof says about your deck.
Upload your pitch deck and get VC-grade analysis in 60 seconds — free.