Evidence-based scoring is the practice of assigning a score to a person, company, or deal based on documented, verifiable facts about what they actually did, not on impression, pedigree, or a good story. Every point on the score is traceable to the specific piece of evidence that earned it.
The mechanics are simple even when the underlying judgment is not: define the criteria that matter for the decision, collect real evidence against each one, and let that evidence, not a reviewer's overall impression, set the score. If a founder claims they scaled a team from 5 to 50, the evidence is the org chart history and the outcomes those hires produced, not the line in the deck.
This differs from a gut-feel rating in one important way: it produces an audit trail. Anyone can trace a final number back to the exact facts that supported it, which is what makes the score defensible in a partner meeting or a hiring committee months later.
The research backing this approach keeps getting stronger. Analysis compiled by Second Talent's 2026 AI recruitment statistics found that cognitive ability tests, one of the most rigorously evidence-based assessment methods available, have a meta-analytic validity of r = .51 for predicting job performance, well above unstructured interviews or resume screening alone. The same body of research frames the real question for any scoring tool as not "does it use AI" but "does it predict the outcome it claims to predict," which is the entire premise of evidence over narrative.
A parallel 2026 guide from Cogn-IQ on AI hiring assessments makes the same point from the HR side: tools with the weakest evidence base, like scoring candidates on facial expression or vocal tone in video interviews, are exactly the ones under the most scrutiny in 2026, while tools tied to demonstrated, checkable work hold up. The pattern is consistent across venture diligence and hiring: the closer a score sits to real, checkable evidence, the more it holds up under challenge.
Evidence alone is not enough without a shared standard for how much each piece of evidence counts. That is the job of an evaluation rubric: it defines the categories, the weights, and what "strong evidence" looks like in each one, so two reviewers scoring the same founder or candidate land on the same number. Our own founder scoring approach and the broader due diligence process both run on this same foundation. For the mechanics of how we tie every score back to its source, see how we tie every score to a line of evidence, or request a demo to see it against a real founder or candidate.
No. The score organizes and weighs the evidence. A human still makes the final call, but they make it with the full record in front of them instead of a summary written from memory.
Anything checkable: shipped work, public writing, verified outcomes, prior hiring or team-building results, reference-confirmed claims. The bar is that it can be independently verified, not that it came from the subject themselves.
It reduces the specific bias that comes from pattern-matching on pedigree or likability, since the score is anchored to facts rather than impression. It does not remove the need for a well-designed rubric, which still has to be checked for fairness.
Bring a founder or a candidate. We'll show you the score and the exact line of evidence behind every point.