Key Takeaways
- Health scores are valuable for scanning portfolios at scale but dangerous if treated as the sole source of truth
- Most scores rely on lagging indicators like usage and tickets, missing executive changes, strategic shifts, and quiet dissatisfaction
- Green scores create false confidence—the measurable signals can be positive while the customer is already planning to leave
- Treat scores as a starting point for investigation, and layer in qualitative human insight through manual sentiment checks
The dashboard says the customer is healthy. Green across the board. Good usage numbers. No open support tickets. NPS score looks fine.
Three weeks later, they churn.
Every customer success team has a story like this. The health score was green, and then suddenly it wasn't—except by the time it turned red, the customer had already made their decision. The score was lagging so far behind reality that it provided no warning at all.
Health scores are supposed to help you predict problems before they happen. When they work, they're invaluable—a way to focus attention on customers who need it, before it's too late. When they don't work, they're dangerous—false confidence that lets at-risk accounts slip through. Effective customer education can help reduce churn, but it won't show up in traditional health score metrics.
Understanding what health scores can and can't tell you is the difference between using them wisely and being misled.
What health scores actually measure
Most health scores combine several types of signals:
- Product usage. Are they logging in? How often? Which features? How does their usage compare to similar customers?
- Support interactions. How many tickets? What severity? How quickly resolved? High volume might indicate problems—or an engaged customer.
- Relationship signals. When did you last talk to them? How responsive are they? Do they attend events?
- Business outcomes. Are they achieving what they bought the product for? Hardest to measure but most meaningful.
- Contract data. When does renewal happen? Have they expanded or contracted? Paying on time?
Different models weigh these differently. Some are simple—usage above threshold equals healthy. Others are sophisticated—machine learning trained on historical churn. But they all share a limitation: they can only measure what's measurable.
What health scores tell you
When built well, health scores provide useful signal:
- Early warning on usage drops. If a daily user becomes weekly, something changed. The score catches this before you'd notice manually.
- Patterns across the portfolio. With hundreds of customers, you can't monitor each one closely. Scores let you scan for problems at scale.
- Benchmarking against peers. Is this customer's usage normal for their size and segment, or unusually low?
- Tracking trends. Is health improving or declining? Even if the absolute score is hard to interpret, direction tells you something.
- Expansion opportunities. High scores often correlate with expansion readiness.
These are real benefits. A CS team with good health scores is more effective than one flying blind.
What health scores miss
Here's where it gets dangerous.
Executive changes. Your champion leaves. A new VP comes in with different priorities or a competitor relationship. This doesn't show up in usage data until it's too late.
- Strategic shifts. The company changes direction. The initiative your product supported gets deprioritized. They're still using the product out of habit, but renewal isn't in their plans.
- Quiet dissatisfaction. They're using the product, but they're not happy. They've stopped complaining because they've stopped caring. They're just waiting out the contract.
- Competitive threats. A competitor has been running a stealth sales process. Your contact didn't mention it because they're the ones pushing for the switch.
- Budget pressure. Economic conditions change. They're cutting costs across the board. Your product might be valuable, but not valuable enough to survive the budget review.
- Poor handoffs. Sales to CS transitions that lose critical context create accounts that were never set up for success from the start.
- Relationship decay. You haven't talked to them in months. The score looks fine because usage is fine. But you've lost the relationship, and when a problem arises, they won't give you a chance to fix it.
None of these show up in typical health score inputs. They're knowable—through conversation, through paying attention, through maintaining relationships—but they're not measurable.
The false confidence problem
The biggest risk of health scores is that they create a sense of security that isn't warranted. When the score is green, it's easy to deprioritize that account. Focus on the red ones. The green accounts are fine—the data says so.
But the data is incomplete. The green score means the measurable signals are positive. It doesn't mean the customer is healthy. It doesn't mean they'll renew. It just means the things you're measuring look okay.
This false confidence is why healthy-looking customers sometimes churn without warning. The warning signs were there—just not in the data the health score was watching.
Making health scores useful
The answer isn't to abandon health scores. It's to use them for what they're good at while compensating for what they miss.
- Treat scores as one input, not the answer. Combine with human judgment, relationship knowledge, and direct conversation.
- Investigate green accounts too. Don't only dig into red accounts. Periodically check in with healthy-looking customers. Make sure the score reflects reality.
- Watch for score changes, not just absolutes. A customer dropping from 90 to 80 might be at more risk than one stable at 70. Direction matters.
- Include qualitative signals. Build processes that capture what scores miss—champion changes, strategic shifts, competitive activity. This requires CSM input.
- Validate against outcomes. Track whether scores actually predict churn. If healthy-scored customers are churning at meaningful rates, your score isn't working.
- Don't let scores replace relationships. The best early warning system is a strong relationship where the customer tells you when something's wrong. Consider using AI roleplay for customer success to help CSMs practice difficult conversations before they happen.
Operationalizing human insight: Implement a "sentiment check" alongside automated scores. Require CSMs to manually flag account status (Green/Yellow/Red) based on their last interaction, regardless of usage data. If automated score is Green but CSM flags Yellow, that discrepancy triggers a review. This forces soft intelligence into the formal record.
The human layer
Health scores work best as a starting point for human attention, not a replacement for it.
The score says this customer might be at risk. Great—now a CSM investigates, has a conversation, figures out what's really going on. The score directed attention; the human figured out the truth.
The score says this customer is healthy. Also useful—but the CSM still checks in periodically, maintains the relationship, asks questions that surface what data can't see.
The score is a tool. The CSM is still doing the work. Teams that over-rely on health scores—treating green as "don't worry" and red as "panic"—miss the nuance that determines real outcomes.
Health scores will get better over time. Machine learning will find patterns humans miss. More data sources will be incorporated. The predictions will improve.
But they'll never be complete. Customers are humans making decisions for complex reasons, many of which won't show up in any data system. The score can tell you about behavior. It can't tell you about intentions, priorities, relationships, or politics.
Use health scores. Build them well. Watch what they're telling you.
And keep talking to your customers. That's where the real signal is. For deeper insight into how knowledge access affects customer interactions, explore how custom virtual experts can help CS teams respond faster and more consistently.
JoySuite helps CS teams stay informed. Customer context at your fingertips—not just usage data, but conversation history, relationship notes, everything you need to understand what's really happening. Health scores plus human insight, combined with AI that synthesizes across your systems.
