What does Dr. Jonah Tebaa on the Metric That Quietly Stops Measuring What You Think It Measures mean in practice?
Support ticket volume is the metric that quietly stops tracking customer friction once automated deflection tools are introduced, according to Dr. Jonah Tebaa. When automated layers resolve interactions without escalation, visible tickets vanish even as underlying frustration climbs. To restore true visibility into account churn risk, Tebaa proposes a replacement instrument built from three signals: intent-classification tags applied to every automated interaction, a friction score derived from repeat topics occurring within thirty days, and a monthly reconciliation between automated resolution volume and the revenue team, owned by a named individual.
Dr. Jonah Tebaa opens his latest piece with a composite scene — assembled from a recurring pattern rather than any single engagement, with illustrative numbers — that will be familiar to any revenue or customer success leader: a quarterly renewals meeting, a dashboard reading green, and a customer success lead confidently waving off risk on a $340,000 account because ticket volume was down. Three weeks later, the account churned. Nobody in that room had missed anything visible. The instrument they trusted had simply stopped tracking what they believed it tracked.
An Accidental Metric, Not a Designed One
Tebaa's argument centers on a distinction he says most organizations never draw explicitly: ticket volume was never built as a churn predictor. It became one by accident, because for years it was the only behavioral trace a support function generated. Rising contact counts, shifting subject-line clusters, and repeat contacts functioned as an improvised proxy for account friction — useful precisely because it was the only signal on hand. In the composite he walks through, that proxy had been present in roughly sixty percent of the team's confirmed churn saves the prior year, making it their most trusted early-warning input.
His point is that AI-driven support deflection does not just speed up resolution. It changes what generates the raw data in the first place. When a system resolves a contact without human visibility, the exhaust that used to create a ticket disappears with it, even if the underlying friction has not improved at all.
The Case: A Dashboard That Read the Wrong Story
In the composite Tebaa walks through, the account had averaged eleven tickets a month before an AI support layer was introduced. Within three months, AI resolved eighty-eight percent of that account's contacts without escalation, and visible ticket volume dropped to two a month. The dashboard read as improvement. Pulled from the full AI conversation record after the account churned, the real number told a different story: total contact volume had actually risen, to nineteen a month. The friction had gotten worse, not better — it simply stopped producing a ticket anyone counted. Tebaa frames this as the metric changing jobs without anyone updating its definition: it had quietly begun measuring AI throughput instead of customer friction.
What He Recommends Instead
Tebaa's prescription is not to pull back on automation, but to rebuild the evidence stream deliberately before it disappears. He outlines three changes:
- Intent-classification tags on every AI-resolved contact, including ones that never escalate, so volume-by-intent data survives even when human-visible tickets do not.
- A friction score derived from repeat-topic-within-30-days per account, calculated directly from the AI conversation record rather than from ticket counts.
- A monthly reconciliation between AI-resolved volume and the revenue team, owned by a named individual, rather than an informal assumption that someone is monitoring the gap.
Properly instrumented, Tebaa argues, those three signals surface a struggling account six to eight weeks ahead of its renewal date — a meaningfully earlier window than a ticket dashboard can produce once deflection sits in the path. He is careful not to present this as a guaranteed outcome or as a promise about any particular team.
The Broader Point
For Tebaa, the underlying lesson extends past support metrics. A ticket was always a proxy for effort and attention, never the thing itself. Leaders who equate a declining ticket count with a healthier account, he argues, will keep getting surprised at renewal regardless of how capable their AI becomes — unless they treat every automated metric as something to re-examine whenever the process generating it changes.
