Customer health score is a composite metric that quantifies how well a customer is succeeding with your product. It aggregates behavioral, engagement, and outcome signals into a single score that predicts the probability of renewal, expansion, or churn. Customer success teams use it to prioritize proactive outreach, identify at-risk accounts before they churn, and spot customers ready for an upsell conversation.
The underlying principle: customers who are actively using the product, achieving measurable value, engaged with your team, and in good contractual standing are likely to renew and expand. Customers showing declining usage, missed outcomes, reduced engagement, and support friction are at risk. The health score makes those patterns visible at scale, so customer success managers (CSMs) do not have to rely on gut feel or manual review of every account to know where to focus.
What Goes Into a Customer Health Score
No two health score models are identical because the signals that predict success vary by product, customer segment, and the specific outcomes you are trying to predict (renewal probability, expansion readiness, churn risk). However, most models pull from four categories of signals:
1. Product Usage
Usage data is typically the highest-weight component in most health scores. Signals include: login frequency, active user count relative to seats purchased, engagement with core features (are they using the features that correlate with value?), and breadth of adoption (are they using one module or the full product?). Usage data is usually sourced from the product analytics system and pulled automatically into the health score model.
The key is identifying which usage behaviors correlate with retention, not just which behaviors indicate general activity. In many products, logging in frequently is less predictive than using a specific feature that delivers the core value proposition. Building the health score model requires analyzing which usage patterns your best customers exhibit — and using those as the signal, not arbitrary activity proxies.
2. Engagement
Engagement signals measure how connected the customer is to your company and team, not just the product. Indicators: responsiveness to CSM outreach (do they respond to emails and attend QBRs?), participation in training or onboarding sessions, attendance at user community events or webinars, and responses to NPS surveys. A highly engaged customer who responds quickly to CSM outreach and attends QBRs is more likely to surface problems before they become churn risks than a customer who is unresponsive.
3. Relationship and Support
The health of the relationship with key stakeholders at the customer account matters for retention. Signals: do you have a strong relationship with the economic buyer (the person who approves the renewal)? Has the main champion or sponsor at the account churned recently (champion departure is one of the strongest leading indicators of account churn)? How many open support tickets exist, and what is the severity? Customers with unresolved high-priority support issues and depleted champion relationships are materially more likely to churn than customers with the same usage but stronger account relationships.
4. Adoption and Outcomes
Whether the customer is achieving the outcomes they purchased the product for is the most meaningful health signal — and usually the hardest to measure systematically. In some products, outcomes are directly observable (a marketing automation platform can measure whether email campaigns are being sent and whether leads are converting; those numbers are the outcome). In others, outcomes are qualitative (a customer is “satisfied” or “achieving ROI”) and need to be captured through periodic success reviews or NPS comments. Whenever outcomes can be made quantitative and tied to customer-specific success criteria established during onboarding, include them in the health model.
Building a Health Score Model
Building a customer health score model involves four steps:
Step 1: Define the outcome you are predicting
Most health scores are designed to predict renewal probability. Others are designed to predict expansion readiness. Some are designed to predict product adoption completion (useful for early-lifecycle scoring). Define the outcome first, because it determines which signals you need and how you weight them. A model predicting expansion readiness is built differently than a model predicting churn risk: expansion-ready customers are those with high adoption depth, high engagement, and specific usage patterns suggesting they are constrained by their current tier.
Step 2: Identify your best and worst customers
Pull your historical renewal/churn data and create two groups: customers who renewed (healthy outcomes) and customers who churned (negative outcomes). For each group, pull available data on product usage, engagement, support volume, and relationship health at a point 90 days before their renewal date. What patterns distinguish the two groups? Those patterns become the basis of your health score model.
Step 3: Select and weight signals
Based on the pattern analysis, select the signals most predictive of your outcome and assign relative weights. A common starting framework: 40% product usage, 30% engagement, 20% relationship and support health, 10% outcomes or custom signals. Adjust the weights based on what your historical data shows correlates most strongly with your outcome — no default weighting is universally correct.
Step 4: Set thresholds and review cadence
Convert the composite score to a health tier: green (healthy, minimal intervention needed), yellow (at risk, CSM should check in), red (critical, escalated attention required). Set the thresholds so that the distribution of your current customer base across tiers is actionable — if 80% of accounts are red, the thresholds are miscalibrated. A functional model typically produces 60-70% green, 20-25% yellow, and 5-15% red accounts. Review the model quarterly to check whether score movements are actually predicting outcomes.
Common Health Score Mistakes
Using activity instead of value signals
A customer logging in daily is not the same as a customer achieving value. Health scores that weight generic activity (logins, page views) without considering whether the customer is using the features that drive outcomes can show false positives — customers who appear healthy based on activity but are actually not achieving results and are quietly evaluating alternatives. Tie usage signals to value-delivery features, not just any activity in the product.
Building the model without historical validation
Building a health score model without testing it against historical churn and renewal data is building on assumption rather than evidence. Before deploying a model, backtest it: apply your proposed signals and weights to customers 90-180 days before their renewal date and check whether the score correctly predicted their outcome. A model that correctly predicts churn 70% of the time is useful; a model that is right 40% of the time is not better than guessing and may mislead CSMs into prioritizing the wrong accounts.
Setting it and forgetting it
Customer behavior patterns change as your product evolves, as customer segments shift, and as the competitive landscape changes. A health score model built in year one may be using signals that are no longer predictive in year three. Review the model at least quarterly: check whether yellow and red accounts are actually churning at higher rates, and whether the signals you weighted most heavily are still the most predictive. Refine the model based on what you observe.
Operationalizing the Health Score
A health score is only valuable if it drives action. The operational workflow:
- Weekly red account review: CSM managers should review all accounts that moved into red in the past week and assign specific interventions: an executive check-in, a product deep-dive session, escalation to the support team, or a success plan review.
- Yellow account proactive outreach: CSMs should have a standard playbook for yellow accounts that is not waiting for a scheduled QBR — a direct email, a check-in call, or a specific request to discuss what value the customer is seeing and what is getting in the way.
- Expansion triggers: Accounts showing high health scores combined with high feature adoption and near-limit usage are candidates for expansion conversations. The health score should trigger CSM workflows for these accounts as well, not just for at-risk ones.
Customer health scores become most powerful when they are embedded into the CSM workflow through the customer success platform (CSPs like Gainsight, Totango, or ChurnZero) rather than living in a spreadsheet that someone has to manually update. Automated score calculation, alerting, and playbook triggering based on score thresholds transform health scoring from a reporting tool into an operational system.