What is a customer health score, and how does a CRM calculate one?

Renewal conversations go badly for a predictable reason: the warning signs were sitting in the data weeks earlier, but nobody was looking at the right combination of them. A customer health score exists to fix that — to compress usage, support, and engagement data into one number a rep or account manager can check without digging through every activity log.

What is a customer health score?

A customer health score is a composite metric, usually on a simple scale like 0–100 or red/yellow/green, that estimates how healthy a customer relationship is right now. “Healthy” generally means likely to renew, likely to expand, and unlikely to churn — the opposite of the warning signs tracked by deal rot on the sales side, but applied after the deal is won.

It is not one metric on its own. A single low number — say, low product usage — might mean nothing if the account is otherwise engaged and paying on time. The score’s value comes from combining several signals so that a real problem stands out from normal noise.

What signals go into the score?

Most health scores draw from four categories, weighted by what actually predicts churn for that business:

A CRM (or a connected customer success platform) pulls these signals in — some automatically, like email engagement and activity tracking, others via integration with your product analytics or support desk — and rolls them into a single score using either a weighted formula or, in more advanced tools, a model trained on which past accounts actually churned.

How do you build one without overcomplicating it?

Teams building their first health score tend to make it too complex, folding in every metric they can access. A better approach:

1. Start with three to five signals you have real conviction predict churn — usually usage and support are the strongest starting points.
2. Weight them by evidence, not intuition. Look at accounts that churned in the last year and see which signals actually moved before they left.
3. Set thresholds that trigger action, not just a colour change — a score dropping into “red” should automatically create a task for the account owner.
4. Revisit the weights periodically. What predicted churn a year ago may not predict it today as your product and customer base change.

What should you do when a score drops?

The score is only useful if it changes behaviour. A dropping health score should prompt a specific, proportionate response — a check-in call for a moderate dip, an escalation to a manager for a sharp one — the same way lead scoring is meant to trigger outreach, not just sit on a dashboard. Pair the score with customer lifetime value so the team knows which accounts are worth the most effort to save, and treat a persistently low score as a churn risk to address directly, not a number to check in on later.