Customer Health Scoring: A Practical Guide for SaaS Teams

A customer health score is a composite metric that predicts how likely a customer is to renew, expand, or churn. It aggregates multiple signals — product usage, support interactions, sentiment, and billing health — into a single score that customer success teams can use to prioritize their time and interventions.

Think of it like a credit score for your customer relationships. Just as a credit score combines payment history, credit utilization, and account length into one number, a health score combines engagement metrics, support data, and satisfaction signals into a single indicator of relationship strength.

The value of a health score is that it transforms subjective gut feelings (“I think this customer might be unhappy”) into objective, data-driven assessments. This matters as you scale. A customer success manager can track 20 accounts based on intuition, but at 200 or 2,000 accounts, you need a system. Health scoring is that system.

The best health scores are predictive, not just descriptive. A good score does not just tell you which customers are unhappy today — it identifies customers who are likely to churn in 30, 60, or 90 days, giving you time to intervene.

The inputs you choose depend on your product and what data you have available. Here are the most common and effective inputs, roughly ordered by predictive power:

• Login frequency and recency: How often does the customer log in, and when was their last session? Declining login frequency is one of the strongest churn predictors. A customer who logged in daily but has not logged in for two weeks is at risk.
• Core feature usage: Are customers using the features that deliver your product’s primary value? Track usage of 3–5 core features. Customers who engage with multiple core features are stickier. Feature breadth often matters more than total session time.
• Support ticket volume and sentiment: A sudden spike in support tickets, especially unresolved ones, signals frustration. Conversely, zero support interaction for an enterprise customer might indicate disengagement rather than satisfaction — they may have stopped trying.
• NPS or CSAT scores: Direct sentiment data from surveys. A detractor (NPS 0–6) is a churn risk. A recent score drop from promoter to passive warrants attention.
• Payment history: Failed payments, past-due invoices, or requests for discounts can signal financial stress or declining perceived value.
• Contract and expansion signals: Approaching contract renewal, recent downgrades, or declining seat count are structural indicators. Conversely, adding seats or upgrading is a strong positive signal.

There are three common approaches to building a health scoring model, each with increasing sophistication:

1. Simple weighted average (recommended starting point). Assign each input a score from 0–100 and weight them by importance. For example:

• Login frequency (30% weight): 0–100 based on logins per week relative to their historical average
• Feature usage (25% weight): 0–100 based on number of core features used
• Support health (20% weight): 100 minus penalty for open or escalated tickets
• NPS/CSAT (15% weight): Mapped from survey score to 0–100
• Payment health (10% weight): 100 for current, 50 for late, 0 for past due

Multiply each score by its weight and sum for the total. This approach is transparent, easy to explain to the team, and quick to implement.

2. Traffic light system (red/yellow/green). Instead of a numeric score, classify each input as red, yellow, or green based on thresholds. If any input is red, the customer is red. If no inputs are red but any are yellow, the customer is yellow. Otherwise, green. This is simpler but less nuanced — it works well for smaller customer bases or teams new to health scoring.

3. Machine learning models. Train a model on historical data to predict churn probability. The model discovers which inputs and combinations are most predictive without you specifying the weights. This approach requires sufficient historical data (typically 1,000+ churn events) and data science resources, but produces the most accurate scores.

A health score is only valuable if it drives action. Here is a framework for turning scores into interventions:

Green (healthy) customers:

• Focus on expansion opportunities — these customers are satisfied and may be ready for upselling.
• Request testimonials, case studies, or referrals.
• Share advanced features or beta access to deepen engagement.
• Maintain light-touch engagement (quarterly business reviews, product update emails).

Yellow (at-risk) customers:

• Trigger a proactive outreach from the customer success manager within 48 hours.
• Schedule a call to understand what has changed and whether there are unmet needs.
• Offer a training session, share relevant help documentation, or provide a personalized product walkthrough.
• Monitor closely for further decline toward red.

Red (high-risk) customers:

• Escalate to a senior customer success manager or manager immediately.
• Develop a specific retention plan with a timeline and measurable goals.
• Consider offering concessions (discount, free month, feature unlock) if appropriate, but focus first on understanding and solving their underlying problem.
• Set a decision point: if the customer does not improve after your intervention period, prepare a graceful offboarding experience.

To scale health score interventions beyond what your team can handle manually, build automated workflows triggered by score changes:

Automated alerts: Configure alerts in your customer success platform (Gainsight, Vitally, Totango, or even a custom solution) to notify the assigned CSM when a customer drops below a threshold or declines by more than a set amount within a period. For example, “Alert CSM when health score drops below 50 or declines by more than 20 points in 14 days.”

Automated email sequences: For segments where personalized outreach is not cost-effective (typically SMB and self-serve), trigger automated re-engagement emails when the health score drops. These emails should offer help, share relevant tips, and include a clear call to action. Be careful not to spam — limit to 2–3 emails per decline event.

In-app interventions: Use the health score to trigger in-app experiences. For declining users, surface contextual help, feature recommendations, or a prompt to contact support. For healthy users, show upgrade opportunities or new feature announcements.

Regular reviews: Schedule weekly or bi-weekly team reviews of health score distributions and trends. Look at the aggregate: are more customers moving to red? Is a recent product change causing widespread score declines? These patterns are often invisible at the individual customer level but clear in the aggregate data.

Start simple. A spreadsheet-based health score with manual CSM follow-up is better than no health scoring at all. Automate and sophisticate as your customer base grows and the ROI justifies the investment.