A SaaS company can report a net revenue retention figure that would make any board happy and still be losing a majority of its customer base underneath it.
That’s the uncomfortable pattern venture investor Cassie Young laid out in her own newsletter last October, drawing on a scaling stretch from early in her career. At Sailthru, she watched net retention run well over 120% while gross retention sat at 60%, not a typo, just a fast-growing segment of the customer base papering over a much larger group quietly walking away. Net retention, she argues, can never be read on its own. It has to be checked against gross retention, or a company ends up celebrating a number that’s hiding the opposite story underneath it.
Two Retention Numbers, Two Different Stories
Net retention and gross retention answer different questions, and 2026 is the year the gap between them stopped being a rounding error.
Expansion is what net retention factors in: upsells, seat growth, add-on modules. A handful of accounts growing fast enough can offset a much larger number of accounts quietly shrinking or leaving, and the blended number still looks fine on a board slide. Gross retention strips expansion out entirely. It only measures what a company kept. It’s a much harder number to hide behind.
Kyle Poyar’s research with ChartMogul puts real figures on how wide that gap has become, particularly for AI-native products. Median gross revenue retention across AI-native companies sat at just 40%, against 82% for traditional B2B SaaS, and products priced under $50 a month saw gross retention fall to 23%. Young describes an industry-wide reckoning on gross retention as all but inevitable. Her argument: a company can be signing new logos and growing headline revenue while the base it already sold to is disappearing beneath it, and net retention alone won’t show that until it’s too late to matter.
The uncomfortable part is the mechanism itself, not just the AI-native segment where it shows up starkest. Any business running a healthy expansion motion can post strong net retention for a while even as its underlying base erodes. The signal that would catch it early doesn’t live in the revenue line at all.
The Signal Was Always There, Just Not Where Anyone Was Looking
Cancellation decisions don’t happen the moment someone clicks the cancel button. By the time Young’s Sailthru cohort showed up as a 60% gross retention figure, the accounts behind that number had almost certainly been disengaging for months. A revenue line only reports the outcome. It says nothing about the weeks of quiet drift that preceded it.
That’s the gap Dovetail Software has been watching customer success teams run into for years, long before churn had its own AI-driven news cycle. “Product analytics have always fed account-health scoring,” Dovetail Software says. “That part isn’t new.”
What Dovetail Software argues has actually changed is narrower and more specific: understanding the qualitative context behind an account, at scale, across every account in a book of business, without a CS manager having to read a single call transcript to find it. Before AI made that tractable, a CSM with a hundred accounts had product usage numbers and a queue of tickets to react to. Whatever a customer was actually saying, in calls, in support threads, in the language they used when something started going wrong, sat unread until someone had a reason to go looking.
Reading Accounts the Way Product Teams Already Read Data
The comparison Dovetail Software draws is to quantitative product analytics a decade ago. Before dashboards and warehouses made usage data self-service, product teams waited on whoever controlled the reporting. Once that data became something anyone could query directly, teams stopped waiting for permission to look at it.
Customer success is on the other side of that same shift now. “CS can now read the qualitative health of accounts the way they’ve always read quantitative signals,” Dovetail Software says. “That reshapes where they spend their attention.”
The distinction matters because the accounts that look fine on a usage dashboard are exactly the ones the gross-retention numbers keep flagging as a blind spot. A team logging in regularly and running one report a week is generating clean usage data and heading for the exit at the same time. The gap between “still logging in” and “still getting value” is invisible to a login counter and fully visible in what the account is actually saying, if anyone’s reading it continuously rather than during a quarterly QBR prep.
Agents Change What Customer Success Spends Its Time On
None of this compounds unless someone, or something, is actually keeping up with it across every account, every week.
Dovetail’s Sun’s Out launch describes its AI Agents as the teammates you never had the headcount for, running continuously against every account rather than waiting for a CSM to open a transcript. Applied to churn specifically, that means an agent surfacing a shift in how an account is talking about the product, before that shift shows up as a login drop or a downgrade request.
“Agents surface this continuously, so CS spends time acting on insight instead of assembling it from call recordings and CRM exports,” Dovetail Software says.
That’s a different allocation of a CSM’s week than the one most retention playbooks assume. The traditional model has a CSM manually triaging a hundred accounts, reading what they can, and reacting to whatever surfaces loudest. An agent-monitored model has the qualitative read already done, continuously, so the CSM’s time goes to the accounts where something is actually shifting, not to the manual work of finding out which accounts those are.
Gross Retention Is a Symptom. The Signal Gap Is the Disease.
The retention conversation happening across SaaS right now keeps landing on the same conclusion from different directions. Net retention can hide a leaking base. Cancellation decisions get made weeks before they show up anywhere a dashboard would catch them. And the qualitative data that would catch it earlier has, until recently, been too expensive to read at the scale a real book of business requires.
Fixing the gross-retention number without fixing the underlying signal gap just delays the same problem to next quarter. The companies treating this as a signal-density problem rather than a metrics problem are the ones positioned to catch the shift 60 days out instead of reading about it in next quarter’s board deck.
