Reconstructed first-round book from the research data, not a scrape of today's website. Public portfolio pages over-report winners, and this graph has the same public-data bias. Investor edges are essentially Crunchbase through late 2015, plus Y Combinator.
Research book · 3-year follow-on
Subtraction Capital vs Kairos
Not ranked
Whole-book follow-on hit rate 80.0% (4 of 5 labeled deals). 95% Wilson interval 37.6%–96.4%. Vintage-weighted base rate 52.9%. This is not a fund return.
Not ranked. A book needs at least 15 labeled deals with a vintage base rate. This one has 5.
Vs same-vintage base rate
+27.1 pp
80.0% on 5 deals vs 52.9%
Vs median VC book
+7.1 pp
80.0% on 5 deals vs 72.9%
Vs top-quartile VC
-8.9 pp
80.0% on 5 deals vs 88.9%
Vs Y Combinator
+7.4 pp
80.0% on 5 deals vs 72.6%
Ranked when at least 15 labeled deals have a vintage base rate. Rank is the average-rank percentile of shrunk excess (20 pseudo-deals at that book's vintage base rate) among eligible VC books that clear the same minimum. 100 is best. The whole book is never compared to Kairos's top-50 hit rate.
What Kairos would have said
Scores are the model’s percentile at each company’s first observed funding event, not at a later round. That matches this fund’s entry only when the fund was in the first-round syndicate. A deal is flagged when its percentile clears that vintage’s top-50 cutoff. The whole book is not compared to the top-50 hit rate.
Deals with a score
5
Mean model percentile
70
Flagged deals
0no labeled flagged deals
Companies
5 labeled follow-on deals, 1 matched without an observable label (dropped, not counted as failures).
| Company | Vintage | Follow-on | Kairos percentile | Flagged |
|---|---|---|---|---|
| Bloodhoundbloodhound.com | 2011 | Yes | 89 | — |
| Medlertmedlert.com | 2012 | Yes | 78 | — |
| Propellerusepropeller.com | 2013 | No | 71 | — |
| eSharesesharesinc.com | 2013 | Yes | 67 | — |
| HealthLoophealthloop.com | 2013 | Yes | 47 | — |
| Prompt.lyprompt.ly | 2013 | Unobserved | — | — |