Customer Engagement Score: the Churn Number You See Coming
A customer engagement score is churn's leading indicator: one number per account, computed from the events that predict renewal. Here's what goes in, how to weight it, and the back-test that keeps it honest.
Churn is a lagging indicator — by the time it shows up in the dashboard, the customer already decided months ago. A customer engagement score (CES) is the leading version: one number per account that says how alive the relationship is right now, computed from what the customer actually does in the product. Sales teams renew against it, success teams triage by it, and the best pricing conversations start from it.
CES = Σ (event weight × event count), normalized 0–100Events that predict renewal get weight; logins alone predict nothing.
What goes in the score
The mistake every first attempt makes is measuring activity instead of value. Logins, page views and session length are attendance, not engagement. The events worth weighting are the ones that mean the customer is getting the outcome they bought:
- Core value events — the action the product exists for: reports exported, campaigns sent, invoices processed, identities verified.
- Breadth signals — seats active as a share of seats paid, teams or departments touching the product, features adopted beyond the landing one.
- Integration depth — connected data sources, API calls, workflows wired into other systems. The founders we interview say it in different ways, but rip-out cost is retention.
- Growth motions — invites sent, workspaces created, usage approaching plan limits — the events that precede expansion revenue.
Building one that works
- Find the events that separated renewals from churns. Pull a year of history and compare what churned accounts did in their last 90 days against what renewing accounts did. The difference is your event list.
- Weight by predictive power, not intuition. If accounts using two integrations renewed at twice the rate, integrations earn a heavy weight — whatever the product team’s favorite feature is.
- Normalize by plan and cohort. A 10-seat startup and a 500-seat enterprise have different healthy baselines; score against the account’s own expected profile, or big accounts will always look engaged.
- Wire it to action thresholds. A score is furniture until a falling one opens a success ticket and a rising one near plan limits alerts sales to the expansion.
The honesty rule: a CES is a prediction machine, so score it like one. Every quarter, check whether last quarter’s low scores actually churned and high scores actually renewed or expanded. A score nobody back-tests drifts into astrology with a dashboard.
What it feeds
Engagement scores earn their keep downstream. They’re the early-warning input to churn before it happens, the qualifier that tells sales which accounts can absorb a price increase, and the missing variable in lifetime value — because average lifetime is just engagement integrated over time. Companies with genuinely expansion-driven models — the 120–150% net-retention businesses in our interviews — are running some version of this arithmetic whether they call it a CES or not: their pricing meters are engagement scores with a billing API.
Start simple: five weighted events, one spreadsheet, back-tested quarterly, beats an unvalidated ML model every time. The rest of the metric system it plugs into lives at the SaaS metrics hub.
SourcesRetention and expansion patterns as disclosed across Nathan’s founder interviews in the GetLatka archive.