Verifying every departure date, not just the easy ones
A gap in our run procedure let a few exit dates slip through with the wrong date. Here's what we found and the check we added so it can't happen quietly again.
One of the harder things about tracking VC portfolios from public pages is dating a change correctly. When a company drops off a firm’s portfolio page, the interesting question isn’t just that it happened — it’s when. And “when” is easy to get subtly wrong.
What we found
Our run procedure re-dates a detected change when there’s public evidence that the real event happened on a specific date — a firm’s own announcement, for example — rather than just recording the date our scraper first noticed it. That re-dating step had a blind spot: it would confirm a date from the firm’s portfolio page, but in a handful of cases it never opened the second source that would have corrected it. The result was a small number of departures and one addition carried with a detection date when a better, source-verified date was actually available.
None of these were wrong in the sense of “this company didn’t leave the portfolio” — the change was real. They were wrong in the narrower, and to us more important, sense of being dated to when we noticed rather than to what the evidence supported.
What we changed
We tightened the re-verification pass so it’s exhaustive rather than opportunistic: every departure and addition in scope now gets checked against all available sources before its date is accepted, not just the first source that happens to confirm it. The few affected rows were re-dated and the corrected record was deployed.
Just as important, the check now fails loudly instead of passing quietly. The whole point of a run procedure is that a gap should surface as a flagged discrepancy on the next run, not sit invisibly in the data. A verification step you can’t see failing isn’t really a verification step.
Why we’re writing this down
BeaconStoneVC is an automated, observational record — it will contain mistakes, and we say so plainly in our methodology. What we can promise isn’t perfection; it’s that when we find a systematic gap, we fix the procedure, not just the row, and we tell you about it. This blog is where those changes get written down.