A customer health score is a composite metric that combines signals such as product usage, support history, engagement, and sentiment into a single indicator of how likely a customer is to renew, churn, or grow. It gives customer-success teams an early read on account risk before a renewal decision is made.
Health scores are usually built by weighting several inputs into a score or a color band, such as green, yellow, and red. Common inputs include login frequency, feature adoption, open support tickets, response sentiment, and time since last contact. Because the inputs and weights are chosen by each team, a health score is only as good as the signals feeding it, and a stale or shallow model can give false comfort.
Support data is one of the richest health signals, and it is often the most underused. The pattern of why a customer reaches out, and whether those issues resolve, says a great deal about account risk. Aide, the agentic AI platform for customer experience, treats the structure of support contact as a first-class health signal rather than a side effect.
When customer intents are mapped and tracked, a rising count of unresolved or escalating intents becomes an early, structured churn signal instead of a vague feeling. Automation only counts as resolution once it has been verified, and every automated outcome is recorded and reviewable, so health signals reflect real outcomes rather than contained-but-unresolved conversations. Intents that customers still bring to humans stay tracked alongside the automated ones, giving the people who read a health score a full view of the accounts behind it.
Frequently asked questions
- What goes into a customer health score?
- Common inputs include product usage and feature adoption, support ticket history, response sentiment, engagement frequency, and time since last contact. Each team chooses the inputs and weights that fit its product.
- Is a customer health score predictive?
- It is only as predictive as its inputs. A score built on shallow or stale signals can mislead, while one grounded in real usage and support patterns gives a more reliable early read on renewal risk.