The signal is in the silence: what unlogged activity tells you
Reps log what they are asked to log. The deals that quietly die leave their fingerprints somewhere else entirely.
We looked at ninety days of activity across early access teams and asked a narrow question: at the moment a deal is about to stall, what changes first?
It was not stage. Stage changes late, and often only after someone is asked about it in a pipeline review. It was not sentiment on calls, which stays polite right up until the deal is gone.
What moved first
Reply latency on the buyer side, measured against that buyer's own baseline. Not absolute hours. Some buyers are simply slow. What matters is the ratio against how fast they had been replying two weeks earlier.
- A buyer whose reply time doubles against their own baseline is materially more likely to go dark within three weeks.
- Single-threaded deals show the pattern earlier and more sharply than multi-threaded ones.
- The signal appears well before the CRM stage moves, which is why stage-based forecasting feels accurate right up until it does not.
Why the CRM cannot see it
Because nobody logs a non-reply. There is no activity record for the email that was not answered, no field for the meeting that quietly stopped getting rescheduled. The absence is the data, and absence is exactly what a form-driven system cannot capture.
Your CRM is a record of what people remembered to tell it. That is a very different thing from a record of what happened.
What to do with it
The useful output is not a risk score. Scores get ignored. The useful output is a specific, cheap next action attached to a specific deal: reach the second stakeholder this week, or ask the question that surfaces the blocker.
We are still early here, and the honest caveat is that ninety days across early access teams is a small sample with obvious selection bias. We are publishing the shape of the finding rather than a precise number, and we will update it as the sample grows.
