One number, built from several weak signals, that turns "this account feels off" into something your whole team can act on the same way.
Updated: 24 September 2026
A customer health score is a single rating — typically expressed as a number, a colour, or a level like low/medium/high/critical — that combines multiple signals about an account's engagement and satisfaction into one measure of how likely it is to renew, expand, or churn. Instead of a CSM relying on memory and gut feel for every account in their book, a health score gives the whole team a consistent, comparable way to see which customers need attention first. Below: what typically goes into a health score, how to weight it, and the mistakes that make teams stop trusting one.
Most health scores combine four to six categories of signal:
No single input is a reliable predictor by itself — that's exactly why they're combined. A customer with dropping usage but a recent large expansion might simply be past its peak onboarding phase; a customer with steady usage but three unresolved critical tickets is a different kind of risk entirely. The score exists to weigh these against each other.
There's no universal weighting that works for every business, because risk factors differ by model. A B2B service business with multi-year contracts cares more about executive sponsor turnover and renewal proximity; a transactional SaaS product cares more about weekly active usage and payment failures. The practical approach is to start with reasonable defaults, then adjust the weighting as you learn which factors actually preceded your past churn events — and to keep that weighting visible and adjustable rather than buried in a fixed formula nobody on the team can change.
A health score taken at a single point in time tells you where an account stands today. A health score tracked over time tells you where it's heading — which is usually the more actionable signal. An account sitting at "medium" for six months is a different situation from one that dropped from "high" to "medium" in the last two weeks, even though the snapshot score looks identical. Wherever possible, surface the trend line, not just the current level.
The most common failure mode isn't a badly calibrated score — it's an opaque one. If a CSM sees an account drop to "at risk" and can't see why, two things happen: they can't have a specific, credible conversation with that customer, and after enough unexplained scores they start ignoring the score entirely. A health score only stays useful if every score is explainable down to the individual factors that produced it, and if the team can override or annotate it when they have context the model doesn't. This is the same explainability requirement that applies to churn prediction generally — a health score is really just churn prediction expressed as a single readable number.
A health score that just sits on a dashboard doesn't move any metric. It needs to route into a workflow: risk accounts appear on a prioritised worklist, with a suggested playbook attached, so a CSM's day starts with the accounts that need them most rather than whichever ones happen to have a scheduled call. For the specific plays that follow a health score drop, see reducing customer churn.
Building a health score means pulling usage, billing and support data into one place, which for most cloud CS platforms means sending that data to a third-party vendor. A self-hosted approach calculates the same score without the data ever leaving your own server — worth checking if any of your accounts are in a regulated industry or simply prefer it that way. See how Lonieta keeps customer data on your own infrastructure.
A customer health score is a single rating that combines multiple signals about an account's engagement and satisfaction — usage, payment behaviour, support activity, NPS — into one measure of how likely it is to renew, expand, or churn.
Typically product usage, payment and commercial behaviour, support ticket volume and severity, relationship signals like NPS, and lifecycle stage such as proximity to renewal.
By assigning a weight to each input signal and combining them into a single score or level. There's no universal formula — the right weighting depends on your business model and should be adjusted based on which factors have actually preceded churn in your own customer base.
Usually because the score is opaque — a CSM sees an account flagged at risk but can't see which factors caused it, so they can't act on it credibly and eventually stop checking it. An explainable score that shows its contributing factors avoids this.
Play with the live demo and see exactly which factors drive each customer's score.