Health score

Health scores enable proactive customer success rather than reactive problem solving. Rather than waiting for customers to complain or churn, health scores identify risk early when intervention has maximum impact. Early warning allows customer success teams to engage before customers decide to leave.

Health scores improve team efficiency. Customer success teams manage hundreds of customers; they can't monitor each actively. Health scores automatically prioritise workload: focus intense effort on low-health accounts where intervention can prevent churn, moderate effort on medium-health accounts, and light touch on thriving accounts where intervention isn't needed.

Health scores quantify expansion readiness. Rather than guessing which customers might buy more, health scores identify customers with strong engagement and usage patterns indicating readiness for expansion conversations. This guides sales teams toward highest-probability expansion opportunities.

A health score is a single number, usually 0-100, that bundles together how a customer is actually doing: how often they log in, which features they've adopted, how they talk to support, whether they're growing or shrinking. The point is to predict churn before it happens, so your customer success team knows which accounts to fight for and which are ready to buy more.

Think of it as turning a gut feeling into a number you can sort by. A good CSM can sense who's thriving and who's drifting, but you can't manage hundreds of accounts on instinct. A health score puts everyone in one ranked list: red accounts get the intense effort, green accounts get a light touch, and nobody quietly churns while you weren't looking.

The honest catch: a health score is only as good as the signals feeding it. Build it on the wrong inputs and you'll confidently chase the wrong accounts. So the real work isn't the formula, it's checking which signals actually predicted churn in your own history, and revisiting that as your product changes.

What goes into it

Most health scores blend a few of these:

• Product usage: login frequency and consistency, the strongest leading indicator for most products.
• Feature adoption: are they using the features that make people stick, or just the shallow end?
• Support signals: ticket volume and sentiment (and watch out, this one cuts both ways, see below).
• Engagement: do they reply to your emails, show up to calls, open your release notes?
• Expansion signals: headcount growth, new use cases, climbing usage volume.
• Risk signals: usage dropping off, a champion leaving, a renewal date approaching in silence.

Why it matters

It makes customer success proactive instead of reactive. The whole game is catching a wobble 45-60 days before the customer has mentally decided to leave, because that's when an intervention still works. By the time someone fills in a satisfaction survey, they're often already gone, so weight leading indicators (usage, engagement) over lagging ones (surveys).

It also tells you where the upsell is. High, stable health usually means the customer has genuinely deployed your product and is growing into it, which is exactly the moment to open an expansion conversation rather than waiting for them to ask.

How to apply it

Build the score where your customer data already lives. Say you're running customer success in HubSpot, with deal stages, last-contact dates, and product-usage properties all on the company record. A workflow can roll those into a health score property and flag anything that drops below 40 to the owning CSM. No separate tool, the data and the trigger sit in the same place. If your CRM is lighter, say you're tracking accounts in Close for a sales-led motion, you can still tag at-risk accounts from last-activity and renewal-date fields and surface them in a smart view.

Wire the usage half from your actual product behaviour. Say your real churn predictor is the login-and-adoption pattern, not anything your CRM sees natively. Pipe product events into Amplitude, build a cohort for accounts whose weekly active usage has fallen for three straight weeks, and let that cohort feed the usage component of the score. Leading indicator, caught early.

Stitch the signals together and push the alert to where people work. Say usage lives in your analytics tool and account context lives in your CRM. A Zapier automation can watch for an account crossing the red threshold and drop a message into the right CSM's Slack channel the moment it happens, so a sliding account becomes a task, not a number nobody checked.

Validate against your own history, and don't trust the obvious read. Weight components by what actually predicted churn for you, using past data: look at who churned and trace what their score looked like beforehand. The classic trap is support tickets. A B2B services firm once scored low-ticket customers as healthy, until the data showed low-ticket customers were the ones who churned, because no questions meant no value, while heavy ticket volume meant active, engaged users. They inverted the weighting and the score finally tracked reality. Re-run that check quarterly; a model built a year ago is probably scoring last year's business.

Define health score components based on your business model. For usage-based products, product usage is heavily weighted. For seat-based products, user count and adoption matter more. Build health score around the metrics that actually predict churn and expansion in your business, not generic models.

Weight components based on predictive power. Components that most accurately predict churn should be weighted highest. Use historical data: look at customers who churned and trace what their health score components looked like before churn. Weighting should reflect what actually predicts customer success.

Build health score with leading indicators, not lagging indicators. Customer satisfaction surveys are lagging indicators: customers are already unhappy by the time they answer surveys. Product usage and engagement patterns are leading indicators: they predict churn before it happens. Weight leading indicators most heavily.

Review and adjust health scores quarterly. Customer needs evolve, products change, markets shift. Health score models built a year ago may be outdated. Regularly review which health score components accurately predict current outcomes and adjust weightings accordingly.

A SaaS company built a health score combining: weekly product logins (weighted 30%), adoption of new features (20%), support ticket sentiment (20%), and expansion buying behaviour (30%). Customers scoring below 40 were flagged for immediate customer success intervention. The team discovered customers with low health scores typically didn't churn immediately: they churned 45-60 days later. By intervening at low health score, the team could often identify missing features, misalignments with intended usage, or training gaps. Proactive health score-based intervention reduced churn rate from 8% to 4% within six months, directly improving lifetime value and profitability.

An enterprise software company using SaaS pricing built a health score emphasising feature adoption and user count. When health scores were high, it typically meant the organisation had successfully deployed the software and was expanding user count. Customer success team used high health scores to identify expansion conversations. Rather than waiting for customers to request additional seats, they proactively presented expansion options. Proactive expansion conversations (based on health score signals) closed at 35% rate, compared to 8% for customers contacted randomly.

A B2B services firm built initial health scores around support ticket volume and response times. However, comparing health scores to actual renewal and churn patterns showed this model was inaccurate. Low-ticket customers often churned; high-ticket customers often renewed. Deeper analysis revealed that low ticket volume meant customers weren't getting value (no questions to ask), whilst high ticket volume meant active engagement and value realisation. The company completely inverted the support metric weighting. This adjustment aligned health scores to actual churn prediction, making them useful for customer success prioritisation.