An account churned last quarter, and by the time anyone in sales found out, the CRM said everything was fine — the last logged note was "great call, champion is bought in," dated six weeks earlier. The product data told a completely different story. Three of the account's five seats hadn't logged in for over a month. The one feature that correlated most with renewal, the one their champion had once called a must-have, hadn't been touched since the week before that "great call."

Nobody looked, because nobody on the sales side had ever been given a reason to look there. The CRM said what a rep believed to be true the last time they typed something in. The product told you what was actually happening, in real time, and it had been telling the truth for weeks.

CRM tells you what a rep believes. Usage tells you what's happening.

A CRM field is only as fresh as the last time someone updated it, and reps update it when they remember to, which is generally right before a forecast call and rarely otherwise. Usage data doesn't have that problem. It logs itself, continuously, with no incentive to look better than it is. The gap between those two data sets is where most surprise churn and most missed expansion actually live — not because the signal wasn't there, but because it was sitting in a system sales never had a reason to check.

The signals hiding in your own product

Some of the most predictive signals in any B2B company aren't in Salesforce at all. A seat count creeping past whatever threshold usually triggers an upgrade conversation is a stronger expansion signal than almost anything a rep could type into a note. A power user who logs in daily suddenly going quiet for two weeks is a stronger churn signal than anything on a QBR slide, and it shows up while there's still time to do something about it. A team adopting a specific feature that historically correlates with expansion is worth more than a generic "engagement score," because it's tied to something concrete rather than an aggregate nobody can act on. Multiple new logins from the same account in the same week — new people showing up inside a deal that was supposedly already closed — is an organic expansion signal that usually gets noticed by accident, if at all.

None of this requires guessing. Most companies already have this data sitting in whatever tool tracks product analytics, whether that's Amplitude, Mixpanel, or an internal events table someone on the product team built two years ago and mostly forgot to tell anyone else about.

Why it never gets wired up

The reason isn't technical difficulty, even though that's the excuse that usually gets offered. It's organizational. Product and data teams own the usage data. Sales and CS own the CRM. Nobody owns the line between them, so the two systems sit next to each other, each one blind to what the other already knows, and the disconnect only becomes visible after a deal is already lost or an expansion opportunity already missed. I've sat in postmortems on churned accounts where the usage drop was obvious in hindsight, sitting in a dashboard three people down the hall had been looking at the whole time, in a tool the account team had never once opened.

Wiring it up without a full data team

You don't need a data platform to fix this. You need two or three usage thresholds that your own historical data says actually matter — not every metric available, just the handful that have shown up before, right before an account expanded or right before one churned — piped into whatever system already handles your outbound and account routing. Treat a usage threshold exactly the way you'd treat a funding round or a new VP hire: as a signal that triggers a specific action, routed to a specific person, on a specific timeline.

The account that churned didn't have to be a surprise. The data that would have flagged it three weeks early already existed. It just never had anywhere to go.