Building a Customer Health Score That Actually Predicts Retention

Customer health scores are a foundational tool for any customer success team, yet industry research consistently shows that most models are miscalibrated. They measure how satisfied a customer feels or how actively they use a product, but they cannot reliably answer the question that matters most: will this customer renew? This session, led by Irit Eizips of CSM Practice, draws on an industry-wide survey and years of hands-on consulting experience to diagnose why health scores underperform and how to fix them. The core insight is that customer success is a product of two equally weighted factors—desired outcomes and appropriate experience—and a health score must reflect both sides of that equation. Participants will explore a practical quadrant model that maps relationship strength against value delivery to segment customers by renewal probability, examine real-world examples of relationship scoring and value scoring for both high-touch and low-touch customer bases, and learn how to attach specific CS plays to each quadrant so that a health score becomes an action engine rather than a passive dashboard metric. The session also addresses common pitfalls such as over-reliance on usage data, the danger of the aimless check-in call, and the challenge of measuring outcomes for customers you rarely speak with directly.

Learning objectives

• Explain why health scores built primarily on usage, engagement, and survey data predict customer satisfaction but not retention or growth
• Apply the customer success formula—desired outcomes multiplied by appropriate experience—as the structural foundation for a balanced health score model
• Construct a relationship score that captures depth, breadth, and strength of stakeholder connections across both low-touch and high-touch customer segments
• Design a value score that measures business outcome achievement, methodology adoption, and product consumption rather than feature usage alone
• Use a two-axis quadrant model to classify customers by renewal probability and identify which quadrant warrants which type of intervention
• Map specific CS plays and escalation paths to each health score quadrant so that CSMs always have a data-driven reason to reach out rather than a generic check-in

Key takeaways

• A health score that only tracks product adoption answers whether onboarding succeeded, not whether the customer will renew—both dimensions must be present for the model to be predictive
• Customer success is a multiplication of value delivered and relationship quality; an imbalance in either dimension undermines the entire score
• Customers with strong relationships but low value delivery will typically signal churn in advance, giving CSMs time to intervene; customers with weak relationships and low value are the ones most likely to leave without warning
• Low-touch customers can be scored on value by tracking whether they are consuming your company's methodologies, certifications, and thought-leadership content—not just product features
• Health scores should be designed to tell CSMs what to do next, not just to flag risk; granular sub-scores for relationship and value enable targeted plays rather than vague escalations
• Risk triggers—specific conditions with a high independent probability of causing churn—should be tracked separately from the overall health score and tied directly to defined playbooks and escalation protocols
• Sharing a value score with customers themselves can drive customer accountability and turn the health score into a tool for co-creating success plans rather than an internal-only metric