Seven years, zero tickets, one cancellation: the churn signal no health score catches

SaaStr says it canceled Notion last week after seven years, and by its own account the relationship produced zero support tickets, zero feature requests and zero escalations, with one teammate invited to present at Notion's user conference ^{[1] [2]}. That is the part worth sitting with: the account that churned was, on every input a customer success team actually collects, a reference customer.

The history is ordinary. Notion arrived at SaaStr around 2019, when the team was bigger, and the staff meeting moved off a Google Doc into Notion, with every person maintaining a dashboard ^[3]. Usage then narrowed over the years until one job was left, the Monday staff meeting, which by the end was the whole job ^[4]. Meanwhile a teammate built an internal system called 10K, which started as a dashboard, became the AI head of marketing, and now functions as AI VP of Revenue and Finance ^[5]. The Monday meeting migrated to 10K because 10K already held the revenue numbers, campaign data, collections queue and pipeline, and it became the source of truth ^[6].

Note what did not happen. According to SaaStr there was no incident, no outage, no pricing dispute, no bad renewal conversation and no competitor in the account, and there was no bake-off for anyone at Notion to lose ^{[7] [8]}. Usage decay would have shown up on a dashboard, but by the time usage reached zero the replacement had been running for months, so the signal arrived after the decision was already irreversible ^[9]. There was no feature that could have won the account back, because the alternative was an agent wired into the customer's own data ^[10].

The trigger is the detail operators should copy into their own playbooks. The teammate had not logged in for months, and what actually prompted the cancellation was Notion's email telling her she had not logged in in a while ^[11]. SaaStr's read is that re-engagement campaigns assume dormancy is a problem the customer wants solved, when sometimes dormancy is a decision the customer has not gotten around to acting on, and the email is the reminder ^[12]. The advice is not to stop sending them but to know which accounts you are waking up ^[13].

The structural claim follows: B2B health models were built on the premise that unhappy customers tell you and quiet customers are fine, and agentic replacement inverts that, because low touch usually means the product does one narrow job, and one narrow job is exactly what an agent can absorb ^[14].

SaaStr then points the same lens at itself. When 10K was connected to Brex and QuickBooks, it found two customers still paying $300 a month for SaaStr Pro, a learning management product SaaStr stopped supporting six years earlier after the person running it left ^{[15] [16]}. Nobody complained, nobody churned, and SaaStr had assumed nobody was being charged at all ^[17]. At list price that is roughly $43,200 collected on a product that was switched off ^[18].

What to watch: whether any vendor ships a health signal that reads absorption rather than dissatisfaction, meaning accounts down to a single workflow and customers standing up internal agents on their own data. Second, whether re-engagement suppression rules start excluding long-dormant enterprise seats. Third, run the billing reconciliation SaaStr ran, because the same blindness pays out in both directions.