Signals in Product usage · 6 signals

The churn signals hiding in product usage.

Mixpanel, Amplitude, your warehouse — what accounts do inside the product, not what they say about it.

Why this source matters

Usage is ground truth. But usage is not value — logins are vanity, breadth without depth is shelfware, and a busy dashboard can belong to an account that's already decided to leave. The signals below separate motion from meaning.

Product usageTypical lead time: 60–120 daysFalse-positive risk: Medium

Core user activity drop

Daily or weekly active users fall 30%+ over 60 days among the account's core user cohort. Usage is the ground truth — everything else is commentary.

Detection logic

Define the core cohort (users active 3+ days/week at peak); track cohort DAU/WAU vs its own peak. Flag sustained 30%+ declines over 60 days.

When it lies to you

Seasonality, holidays, and industry cycles move usage without meaning. Compare against the same period last year, not just last month.

Product usageTypical lead time: 90–180 daysFalse-positive risk: Medium

Feature breadth narrows

The account uses fewer distinct modules or features than it did 90 days ago. Shrinking footprints precede shrinking contracts — customers consolidate onto what works, then question the rest.

Detection logic

Count distinct features/modules touched per account per month. Flag 30%+ contraction sustained over 90 days.

When it lies to you

Completed project phases legitimately narrow usage. Check whether the contraction maps to a finished initiative before flagging.

Product usageTypical lead time: 60–120 daysFalse-positive risk: Medium

Power-user attrition

The top decile of users by activity goes dark. Power users are your internal advocates — when they leave or disengage, the account's immune system is gone.

Detection logic

Identify each account's top 10% of users by activity; alert when 30%+ of them show zero activity for 30 days.

When it lies to you

Role changes, parental leave, and reorgs move individuals. Check HR-level changes (title/department in CRM) before reading account risk.

Product usageTypical lead time: 60–180 daysFalse-positive risk: Low

API and integration usage decline

Programmatic usage — API calls, webhook deliveries, integration syncs — declines steadily. When the pipes go quiet, the product is being unwired from the customer's stack.

Detection logic

Track API/integration event volume per account vs its 90-day baseline. Flag 40%+ sustained declines; sudden drops to zero are migration events, not drift.

When it lies to you

Endpoint migrations and SDK upgrades shift traffic patterns without meaning. Confirm with engineering that the decline isn't a telemetry change.

Product usageTypical lead time: 90–180 daysFalse-positive risk: Medium

Seat utilization falls

Active seats divided by purchased seats drops below half and keeps falling. Shelfware is the easiest line item to cut at renewal — it's pre-justified.

Detection logic

Compute active-seat ratio monthly per account. Flag below 50% with a declining trend; escalate below 30%.

When it lies to you

Hiring freezes and layoffs cut active users without any dissatisfaction. Pair with engagement-per-active-user to separate contraction from disengagement.

Product usageTypical lead time: 90–180 daysFalse-positive risk: Medium

New team activation stalls

A newly added business unit or team never reaches the activation milestone the sales team promised. Failed land-and-expand poisons the whole account's expansion story.

Detection logic

Define activation per use case (e.g., 5 active users + core workflow completed); flag new cohorts that miss it within 2x the normal time-to-value.

When it lies to you

Enterprise onboarding is legitimately slow — some industries take quarters. Calibrate “normal” per segment, not globally.

How to instrument it

Define the core cohort per account and track its DAU/WAU against its own peak; count distinct features touched per month; watch the top-decile power users individually; track API and integration event volume; compute the active-seat ratio monthly. Compare against the same period last year, not just last month.

Common pitfalls

Seasonality moves usage without meaning — calibrate per segment. Telemetry changes masquerade as behavior changes: confirm with engineering before declaring a drop real. And never use global benchmarks; an account's only fair comparison is its own history.

Signals in the other sources

CRM Call transcripts Support tickets Billing & payments Email & Slack

Get your free Account Risk Snapshot

We run signals like these against your 20 riskiest accounts — before you sign anything.