RevOps Strategy
October 10, 2026 · 5 min read · by Ananda Narasimhan
Most customer health scores get built once, during onboarding, and never touched again. They average logins, NPS, ticket volume, and renewal proximity into a single green-yellow-red number, ship it to a dashboard, and stop predicting anything within two quarters. If churn still shows up as a surprise, the score is decoration, not a tool.
The standard model averages five or six inputs into one traffic light. Averaging hides the signal. An account can sit at "yellow" overall while one input — a 60% drop in weekly active seats — is screaming red underneath it. By the time the blended number moves enough to notice, the renewal conversation is already lost.
The fix isn't a more complete formula. It's separating leading indicators from lagging ones and weighting them by how far out they actually predict churn, not by how easy they are to pull from your CRM.
Across the accounts we've audited, three signals consistently flag churn risk 60-90 days before the renewal date forces the conversation: seat utilization trending down over two consecutive months, a champion leaving or changing roles inside the account, and a drop in usage of one core workflow specifically — not total logins, which often stay flat even as engagement with the product's actual value is already dying.
NPS and support ticket volume are lagging indicators. They describe what already happened, not what's about to. Keep them visible for context, but don't let them drive the score.
Teams that already have a health score usually don't need a new one. They need to stop averaging. The fastest fix is pulling the existing inputs apart into three separate columns instead of one blended number, then checking which column moved first on accounts that already churned. In the audits we've run, seat utilization moves first in roughly two-thirds of churn cases, usually 45 to 75 days before the account actually leaves. Champion changes come second, and they're missed constantly because nobody owns tracking contact changes inside the CRM once a deal closes and the account moves to customer success.
A health score that lives in a spreadsheet gets checked once a quarter, right before it's too late to act on it. It needs to sit where your CS and sales teams already work, updated automatically from product usage data flowing into your CRM. That's a revenue operations problem before it's a customer success one — the score is only as good as the pipeline feeding it, and most teams we audit are missing the product-usage-to-CRM connection entirely. The same data gaps that break a CRM hygiene audit are usually what's quietly breaking the health score too.
We've started building this connection with narrow AI agents instead of broad nightly recalculation rules: an agent that watches specifically for the champion-change and usage-drop patterns, rather than a dashboard that reprocesses everything and buries the real signal in noise. That's the same principle behind our other AI agentic workflows — a specific trigger beats an always-on score every time.
Start with the three leading signals above, not a comprehensive model. Pull two quarters of churned-account history and check whether seat-utilization trend and champion changes would have flagged them 60-plus days out. If they would have, you have a working model before you've touched NPS or ticket data at all. Add lagging indicators afterward, as context for the account team, not as inputs to the score itself.
A health score that predicts churn 60 days out and triggers one specific action beats a "more complete" score that tells the team nothing until the renewal is already at risk.