A customer health score is a single number, built from signals such as product usage, support history, relationship engagement and sentiment, that estimates how likely an account is to renew and expand. The useful ones are weighted by what actually preceded churn in your own history, tested against past renewals, and tied to a named action for every status.
The economics justify the effort. Frederick Reichheld's research, published by Harvard Business School Press, found that "an increase in customer retention rates of 5 percent increases profits by 25 percent to 95 percent" (Reichheld, Loyalty Rules!). A health score is how a revenue team sees that retention risk early enough to act on it.
The retention economics behind a customer health score
Keeping customers is cheaper than replacing them, though the exact multiple is less settled than it sounds. Harvard Business Review puts acquisition at anywhere from five to 25 times the cost of retention, and it frames the range with the caveat that the figure depends on which study you believe and which industry you are in (Harvard Business Review). The direction is consistent. The size varies by business.
Sales leaders have noticed. Gartner's 2025 survey of 243 chief sales officers and senior sales leaders found 73 percent prioritizing growth from existing customers (Gartner). Growth from the base shows up in net revenue retention, and a health score is the early-warning system that protects it.
Customer health score metrics that predict churn
Build from four families of signal. Product usage covers breadth, depth and trend: how many licensed users are active, which core features they use, and whether both are rising or falling. Support history covers ticket volume, severity and time to resolution. Relationship covers executive sponsor contact, champion engagement and meeting cadence. Commercial signals cover payment behaviour, seat utilisation and contract terms. Sentiment, from NPS or a CSM's judgement, sits on top.
Usage deserves the most weight in most software businesses, and the research supports it. A 2024 peer-reviewed study in Industrial Marketing Management, built on 3,959 subscriptions from a European software provider, confirmed the value of usage data for B2B churn prediction (Industrial Marketing Management). Trend matters more than level. An account at 60 percent active seats and climbing is healthier than one at 80 percent and falling.
Calculating a customer health score: weights, bands and decay
Keep the model simple enough to explain in one sentence. Assign points to each signal, cap how many points any group can contribute, weight the groups, and total the result on a 0 to 100 scale. Then cut it into three bands, such as healthy, neutral and at risk. HubSpot's documentation for its customer success workspace works the same way: groups of properties and tracked events add or subtract points, each group has a maximum, scores can decay over time, and thresholds map scores to labels (HubSpot Knowledge Base).
Gainsight's scorecards use the same idea with different vocabulary: each dimension is a measure, and admins set weights for each one (Gainsight documentation). Start with five to seven inputs. Every input you add is one more thing a CSM has to understand before trusting the number, and a score nobody trusts does not get used.
Validate the health score against churn you have already seen
Back-test before launch. Score every account as it looked six to twelve months ago, then compare those scores with what happened at renewal. If accounts marked healthy churned at nearly the same rate as accounts marked at risk, the weights are wrong, and the fix is to reweight toward the signals that separated the two groups. Repeat the test every quarter as the product and customer base change.
Look specifically for signs of buyer regret. Gartner found that 60 percent of technology buyers involved in renewal or expansion decisions regret nearly every purchase they make, and it advises teams to identify accounts displaying behaviours associated with regret (Gartner). Stalled onboarding, a champion who stops attending reviews, and a spike in how-to tickets are all candidates, and they connect directly to the churn reduction playbook.
Running the customer health score inside your CRM
A score lives or dies on what happens when it changes. Store it as a property on the company record, trigger a task or playbook when an account crosses into at risk, and give every band a named owner and a default action. HubSpot creates health score and health status properties automatically when the feature is set up, which makes the score available to workflows and reports; it requires Service Hub Professional or Enterprise (HubSpot Knowledge Base).
Ownership has to be explicit, since renewals and expansion often sit with different teams. The split is covered in customer success vs account management. As CRMs open up to AI assistants, a clean, well-defined health score also becomes one of the first things an agent can act on, a shift covered in our look at headless CRM after Dreamforce 2026.
Customer health score benchmarks and the retention targets they serve
Tie the score to the retention numbers the board watches. The KeyBanc Capital Markets and Sapphire Ventures private SaaS survey reported gross retention declining to 86 percent in 2023 and expected to approach 90 percent in the near term, with net retention remaining above 100 percent (Sapphire Ventures). If your score is working, the gross retention of accounts it rated healthy should sit well above your company average, and the gap should widen as the model improves.
The score tells you which accounts need attention. What you do next decides the outcome, and the best teams choose that next step by account context; that is the idea behind Next-Best Action. If you want help building or rebuilding a health score your renewal team will use, get in touch.
