Well, that was fast. Our post a few days ago highlighted one of the features our AI model takes advantage of. It is one of more than 150 features we use in our platform. Partner Score tracks how good an investment partner is at stage specific investing. We track seed, Series A and B, C and D. We have 3 year, 5 year and 10 year Partner Score models.
With the last post we highlighted the 5 year Series B Partner Scores.
Almost immediately we got feedback from many ecosystem participants that there were some strange hallucinations in the results. Specifically that investors who were able to invest at the Series B and immediately make a unicorn, due to their valuation and check size, were being credited with having invested in a unicorn, even though there was zero value growth.
The goal of Partner Score (and indeed all of our features) is to weight them towards giving credit for outcomes. In this case we were giving credit for writing a check, regardless of the outcomes.
So our correspondents were right. Our reaction was to re-weight the feature and then run it with the new weights. And, no big surprise, we get different results. Thanks to the fact we published a glimpse, and got rapid feedback, we were able to iterate super fast.
There is no embarrassment on our part. We expect to constantly improve each feature over time, often rapidly. Any result is therefore only the current result set and is likely to change over time.
Our models are real time. New funding rounds impact the results on a daily basis.
So that said, here is the result set for the re-weighted 5 year Series B investment stage. Top 100 from over 1000 in the set. SignalRank investors get to see all sets (Seed - Series D) for the 3,5 and 10 year results.