A customer health score tells you which accounts are thriving and which are quietly dying. Get it right, and your CS team saves at-risk accounts before they churn and expands healthy accounts at the right moment. Get it wrong — or worse, don’t have one at all — and you’re flying blind on your most valuable asset: your existing revenue.
The challenge is that a real health score needs data from multiple systems. Most scale-ups either never build one, or build something so basic it’s useless. Here’s how to get it right.
What a good customer health score actually measures
A health score isn’t a single metric. It’s a composite signal built from multiple data sources that, together, tell you whether an account is healthy, at risk, or ready to expand.
The most reliable health scores combine four categories of signal:
Product usage signals
These tell you whether the customer is actually getting value from your product. Key indicators include login frequency and active user count relative to seats purchased, core feature adoption (are they using the features that drive retention?), usage trend over 30/60/90 days (growing, flat, or declining?), and depth of usage (surface-level or embedded in their workflows).
Usage data typically lives in your product analytics platform (Segment, Mixpanel, Amplitude) or your application database. It almost never lives in your CRM.
Financial signals
These tell you whether the revenue relationship is healthy. Key indicators include payment history (on-time, late, failed), plan tier relative to usage (over-provisioned or bumping against limits?), revenue trend (expanding, contracting, or flat), and contract status (approaching renewal, month-to-month, annual).
Financial data lives in your billing system — Stripe, Chargebee, Recurly — and partially in your CRM if someone keeps it updated (they usually don’t).
Support signals
These tell you whether the customer is struggling. Key indicators include ticket volume and trend (increasing tickets often signal frustration), ticket severity (billing questions are routine; “this is broken” is a red flag), time between tickets (a sudden burst is different from steady, low volume), and sentiment in conversations (if your support tool captures this).
Support data lives in Intercom, Zendesk, Freshdesk, or HubSpot Service Hub.
Relationship signals
These tell you whether the human connection is intact. Key indicators include whether the primary contact (champion) is still active, engagement with emails, QBRs, and check-ins, NPS or CSAT scores, and whether there’s been a leadership change at the customer.
Relationship data lives partly in your CRM, partly in your email/marketing platform, and partly in your survey tool. It’s the hardest category to track systematically.
Why most health scores fail
They use one data source
A health score built only on product usage misses the customer who uses your product daily but just had a payment fail. A score built only on billing data misses the customer who’s paying on time but hasn’t logged in for three weeks. A score built only on support tickets misses the customer who stopped complaining — because they gave up and are about to cancel instead.
Each data source tells a partial story. The power of a health score is in combining them.
They’re manual and static
Many companies build health scores in a spreadsheet. Someone pulls data from each system monthly, assigns scores, and produces a report. By the time it’s done, it’s already outdated. A customer that showed warning signs on day 5 of the month doesn’t get flagged until day 30.
They’re too complex to maintain
Some companies go the other direction — they hire a data engineer to build a sophisticated health score model in a data warehouse. It’s accurate and real-time. But it’s also fragile. When a tool changes its API, or a new data source gets added, or the scoring weights need adjusting, the whole thing breaks and the data engineer spends a week fixing it.
They don’t lead to action
A health score is useless if it just sits in a dashboard. It needs to trigger action — alert a CS rep when an account drops into the danger zone, generate an expansion task when an account is thriving, surface the specific signals behind the score so the rep knows what to do.
How to build a health score that works
Step 1: Choose your input signals
Start with the signals you can actually access. You don’t need all four categories on day one. Most companies start with two:
Minimum viable health score: Product usage + financial signals. This alone catches the majority of churn risk and expansion opportunity.
Strong health score: Product usage + financial + support signals. This adds the frustration layer.
Full health score: All four categories. This is the gold standard but requires more data integration.
Step 2: Define “Healthy,” “At Risk,” and “Critical”
For each signal, define clear thresholds. Keep it simple:
| Signal | Healthy | At Risk | Critical |
|---|---|---|---|
| Usage trend (30-day) | Growing or stable | Declined 15–30% | Declined 30%+ |
| Active users vs seats | 70%+ utilisation | 40–70% utilisation | Below 40% |
| Payment status | Current | 1 late payment in 90 days | 2+ late payments or failed charge |
| Support tickets (30-day) | 0–2 routine | 3–5 or severity escalation | 5+ or “cancellation” mentioned |
| Champion activity | Logged in within 14 days | 15–30 days since last login | 30+ days inactive |
The specific thresholds will vary for your product. The point is to make them concrete and measurable, not subjective.
Step 3: Weight and combine
Not every signal matters equally. For most SaaS products, usage signals are the strongest predictor of churn, followed by support signals, financial signals, and relationship signals.
A simple weighting might be: usage (40%), support (25%), financial (20%), relationship (15%). Multiply each category’s score by its weight and sum them.
Don’t overthink the weights initially. A rough weighting that covers multiple data sources will outperform a perfectly weighted score from a single source.
Step 4: Connect it to action
The score needs to trigger something:
- Account drops to “At Risk”: Automatic alert to the assigned CS rep. Include the specific signals that triggered the change.
- Account drops to “Critical”: Escalation to CS manager. Flag for immediate outreach. Include revenue at stake.
- Account scores “Healthy” with expansion signals: Auto-create expansion task. Include which signals suggest readiness (high usage, hitting limits, positive support interactions).
Without this action layer, the health score is just a number on a screen that nobody checks.
The data integration problem (and how to solve it)
Everything above sounds straightforward in theory. In practice, the hard part is getting the data from 4–6 different systems into one place, mapped at the account level, and kept current.
Your options:
Build it yourself. Set up a data warehouse, build ETL pipelines from each tool, create the scoring logic in SQL, build a dashboard, and maintain everything. This works but requires a data engineer and 3–6 months.
Use a customer success platform. Tools like Gainsight or Totango offer built-in health scoring. These are powerful but designed for large CS teams, can be expensive, and still require significant configuration and data integration work.
Use an AI revenue intelligence platform. Eru connects to your billing, CRM, support, and product tools in 5 minutes each. The AI agent maps your data across systems at the account level and surfaces health signals automatically — churn risk accounts ranked by revenue impact, expansion opportunities ranked by readiness, and the specific signals behind each one.
You don’t configure a scoring model. Eru’s AI discovers the patterns across your connected data and surfaces the accounts that need attention and why. It’s a health score that builds itself.
What matters most: start now, refine later
The biggest mistake is waiting until you can build the perfect health score. A basic score using two data sources, launched this week, will save more accounts than a perfect score launched in six months.
Start with what you can access today. Connect more data sources as you go. Refine the weights based on what you learn. The companies that win on retention are the ones that start measuring account health early — even imperfectly — and improve over time.
Frequently asked questions about customer health scoring
What’s the typical cost structure for customer health scoring platforms?
Customer health scoring platforms typically use per-seat pricing ($30–$150/user/month), ARR-based pricing (0.5–2% of monitored revenue), or flat-tier pricing based on customer count. Legacy platforms like Gainsight and Totango use per-seat models that scale steeply as your team grows. Eru uses outcome-based pricing tied to the revenue it helps you protect, so costs stay proportional to results rather than headcount.
How much should I budget for a unified customer health solution?
Most mid-market SaaS companies (200–500 customers) spend $40,000–$120,000/year across multiple point solutions for health scoring, churn prediction, and retention workflows. Consolidating onto a single platform like Eru can reduce total cost of ownership by 40–60% while improving cross-platform visibility. Eru replaces the need to stitch together ChurnZero for health scores, Gainsight for playbooks, and custom CRM formulas.
What are the hidden costs of customer health scoring platforms?
Hidden costs include implementation and onboarding ($10,000–$50,000 for enterprise platforms), ongoing data engineering to maintain integrations (1–2 FTEs), training, and the cost of inaccurate data when systems drift. Eru eliminates most hidden costs by handling cross-system data reconciliation automatically through AI-powered truth graphs, with no separate implementation fee.
What data sources should a customer health score combine?
A reliable customer health score combines four signal categories: product usage (from Mixpanel, Amplitude, or Segment), financial signals (from Stripe, Chargebee, or your billing system), support signals (from Zendesk, Intercom, or Freshdesk), and relationship signals (from your CRM and email platform). Eru connects to 30+ data sources natively and builds unified health scores by reconciling data across all of these systems automatically.
Can I build a customer health score without a data team?
Yes. While most health scoring platforms require significant data engineering to set up and maintain integrations, Eru acts as your data engineer by automatically connecting to your existing tools (Salesforce, HubSpot, Mixpanel, Stripe, and more) and reconciling data across systems. Eru’s AI-powered truth graphs mean you get accurate, unified health scores without writing custom ETL pipelines or maintaining fragile integrations.
Book a free churn audit — we’ll show you which of your accounts are thriving, which are at risk, and what the signals look like when they’re connected.
Book a churn audit →