Your NRR Has Two Engines. One of Them Just Stalled.

A founder I work with runs a $7M ARR product billed per seat. His net revenue retention came in at 112% last quarter, and he led the board update with it. Healthy number. Then I asked him a different question: across his top 20 accounts, were licensed seats going up or down? He didn't know offhand. We pulled it. Logo count was growing. Seats per account had been flat for two quarters and had ticked down in the last one.

His NRR was fine. The thing underneath it had already turned.

The number is an average of two opposite forces

Net revenue retention bundles two engines that move for different reasons. One is price and upsell: customers paying more for the same footprint, buying a higher tier, adding a module. The other is seat expansion: the same customer hiring more people who each need a login. For a decade, seat expansion was the reliable one. You land a team of 10, they grow to 40, your revenue grows with their headcount whether or not you did anything.

That engine ran on an assumption nobody wrote down. More work at the customer means more humans, and more humans means more seats.

Agents break that link. Jason Lemkin put it bluntly this year: if 10 agents do the work of 100 reps, you don't need 100 seats, you need 10. The market took the point seriously. Atlassian dropped 35% after reporting its first enterprise seat decline ever. Salesforce fell 28% on a quarter where revenue still grew, because investors stopped trusting the seat-expansion story behind it. Roughly $2 trillion in software value came off between mid-January and mid-February.

Utilization is the leading indicator your renewal won't be

You don't see this in NRR until the renewal, and by then it's priced. You see it earlier in utilization: active seats divided by licensed seats.

Take one account on 40 licensed seats. A year ago, 34 were active most weeks. This quarter it's 22, because an agent now drafts what eight of those reps used to. The account is happy. Renewal comes up and they right-size to 28 seats. On your books that lands as a 30% contraction inside an otherwise "expanding" cohort, and it shows up the month it's too late to do anything about.

Seat expansion was never loyalty. It was your customer's headcount growth, rented to you. When their headcount stops driving their output, the rent stops too.

Split the number before you forecast off it

Stop projecting next year's NRR off the trailing blended figure. Break it in two. Forecast price and upsell expansion separately from seat expansion, and put a real haircut on the seat line, because that's the one exposed. Then track utilization monthly on your top accounts as the early warning. An account drifting from 85% active to 55% isn't churning. It's telling you the next renewal is a downgrade conversation, and you have two quarters to turn it into an upsell on value the agent can't replace.

The 112% on the board slide isn't wrong. It's just describing an engine that's already losing compression, and the slide won't say so until the rebuild costs more than the warning would have.