{"edition":"3rd edition","observationCutoff":"2026-07-30","editorialRevision":"2026-09-12","canonical":"https://preuve.ai/blog/startup-validation-benchmarks-2026","downloadPath":"/blog/startup-validation-benchmarks-2026/aggregates.json","coverage":"Anonymized data from 6,000+ completed analyses; published statistics use a cleaned, like-for-like comparison set, not unique startups or founders.","provenance":"This is a transcription of selected published aggregates from the July 2026 edition, not raw data, not a new computation and not a verified source export. It preserves the published values without recalculation.","citation":"Vincent, Preuve AI. Startup Validation Benchmarks 2026: 6,000+ Analyses. 3rd edition, observation cutoff July 30, 2026; editorial revision September 12, 2026. https://preuve.ai/blog/startup-validation-benchmarks-2026.","methodology":["The unit is an assessment record. The July edition starts from anonymized data from 6,000+ completed analyses through July 30, 2026 and describes a cleaned, like-for-like comparison set.","The published exclusions are the founder's own test and demo scans, a promotional batch scored by a lighter model, agency scans and investor packages. The remaining scope is free scans, paid deep analyses and follow-up improved iterations.","Risk shares use the published subset of 4,995 records with a stored primary risk label. Rescan figures use 734 selected re-scored iterations, not unique founders. Per-tier and monthly statistics have their own subsets.","All published data is aggregate and anonymized. No individual ideas, founder names or company names are disclosed."],"limitations":["These are model assessment scores and risk labels, not observed startup failure or success rates or calibrated outcome probabilities. No survival follow-up, failure event definition or outcome-validation study is disclosed.","The set is self-selected and includes repeated assessment records. It does not represent all startups. Selection, different tier recipes and model-version changes limit comparisons; aggregate monthly means cannot establish scoring stability.","The exact cleaned-set size and exclusion counts are not disclosed. Monthly and tier denominators, missing-label coverage, pair eligibility, repeat-chain handling, starting timestamp and timezone, and the model-version mix are not fully documented in the public edition.","Rescan differences are uncontrolled associations. They do not establish that rescanning caused improvement, that founders independently completed homework or that business prospects improved. Competitor evidence also enters scoring, so score associations are not independent proof of demand.","Source queries and July aggregate exports are not available with this publication. The numbers cannot be independently reproduced from the materials here. The September editorial correction does not recompute or newly validate the July study."],"rounding":"Percentages are independently rounded. Rounded buckets, verdict shares and risk shares may not sum to 100%, and adding rounded buckets may differ from a separately rounded total. Do not reconstruct exact counts from these percentages.","aggregates":[{"metric":"Go, score 70+","value":17.5,"unit":"percent","subset":"cleaned comparison set"},{"metric":"Caution, score 40-69","value":76,"unit":"percent","subset":"cleaned comparison set"},{"metric":"No-go, score below 40","value":6.4,"unit":"percent","subset":"cleaned comparison set"},{"metric":"Median score","value":54,"unit":"score points, 0-100","subset":"cleaned comparison set"},{"metric":"Mean score","value":56.3,"unit":"score points, 0-100","subset":"cleaned comparison set"},{"metric":"Early-stage primary risk","value":36.9,"unit":"percent","subset":"records with a stored primary risk label"},{"metric":"Go-to-market primary risk","value":25.7,"unit":"percent","subset":"records with a stored primary risk label"},{"metric":"Free mean score","value":55.7,"unit":"score points, 0-100","subset":"free assessment records"},{"metric":"Paid mean score","value":66.6,"unit":"score points, 0-100","subset":"paid assessment records"},{"metric":"Re-scored iterations","value":734,"unit":"iterations, not unique founders","subset":"selected rescan comparison"},{"metric":"Mean rescan score change","value":8.9,"unit":"score points","subset":"selected rescan comparison"},{"metric":"Rescans with higher scores","value":75.2,"unit":"percent","subset":"selected rescan comparison"}]}