What Is a Customer Health Score? How to Build One

A customer health score is a composite indicator, usually a number from 0 to 100 or a red, yellow, or green label, that estimates how likely an account is to renew, expand, or churn. It combines signals such as product usage, support experience, relationship strength, sentiment, and commercial data into one view a customer success team can act on.

If your team finds out an account is unhappy when the cancellation email arrives, you already know the problem a health score is meant to solve. Churn rarely comes out of nowhere. Usage drops, a champion leaves, support tickets pile up, and the signals are usually visible weeks or months earlier. A health score collects those signals in one place and puts someone on the hook to respond.

This guide explains what goes into a score, how to build one step by step, how to test whether it works, and how to set it up in HubSpot.

Why a health score matters

For recurring-revenue companies, keeping and growing existing customers drives net revenue retention, and a health score is the early-warning system behind it. It tells you which accounts need a rescue call, which are healthy enough for an expansion conversation, and where to spend limited customer success time. It also gives sales, finance, and leadership a shared view of renewal risk, which makes your revenue waterfall less of a surprise each quarter.

Health scoring also builds on the customer journey map. The map shows where customers get stuck, and the score tells you which accounts are stuck there right now.

What goes into a health score

Category	Example signals	Type	Typical source
Product usage	Active users, feature adoption, usage trend over 30 to 90 days, seat utilization	Leading	Product analytics
Relationship	Champion present and engaged, executive sponsor, stakeholder count	Leading	CRM, CSM notes
Support	Ticket volume trend, severity, unresolved issues, time to resolve	Mixed	Help desk
Sentiment	NPS, CSAT, tone of conversations, CSM judgment	Lagging	Surveys, calls, notes
Commercial	Days to renewal, payment delays, contract changes, past expansions	Mixed	Billing, CRM

Leading indicators, such as a falling usage trend or a departed champion, change before the outcome does. Lagging indicators, such as NPS, tell you how someone felt after the fact. Weight leading signals more heavily, and don’t build a score on survey results alone.

How to build a customer health score

1. Define the outcome. Decide what the score predicts: renewal, churn within 90 days, or expansion. A single score trying to do all three tends to do none of them well.
2. Look at your own history first. Pull the accounts that churned in the last 12 to 24 months and compare them with ones that renewed. What did the churned accounts look like 90 days before? This beats any generic template.
3. Choose five to eight signals. Customer success platforms often encourage 10 to 15 inputs, but a smaller set you can explain beats a long list nobody trusts.
4. Normalize and weight. Convert each signal to a 0 to 100 scale, then assign weights based on how strongly it predicted churn in your data. There’s no universal weighting.
5. Set thresholds. Decide what counts as green, yellow, and red, and write down what each means.
6. Attach playbooks. Every band needs an action, an owner, and a response time.
7. Backtest, launch, and recalibrate. Check the score against past outcomes before rollout, then review it every quarter.

Score accounts, not individual users. One inactive user in a 40-seat account isn’t a warning sign, but three inactive admins and a missing champion probably are. If you serve different segments, such as self-serve small businesses and enterprise customers, build separate weightings, because the signals that predict churn differ.

A worked example

The weights below are an illustrative starting point, similar to examples published by customer success vendors. They are not a standard. Here’s one account scored on five components:

Component	Weight	Account score (0 to 100)	Points	Why
Usage trend	40%	50	20.0	Usage down 25% over 60 days
Adoption breadth	25%	80	20.0	Uses most core features
Champion and relationship	20%	30	6.0	Champion left; no replacement identified
Support health	10%	70	7.0	A few tickets, mostly resolved
Sentiment	5%	90	4.5	Last NPS response was positive
Total	100%		57.5	Red, if red is below 60

This account would look fine on NPS alone. It scores red because the leading indicators, a falling usage trend and a missing champion, are weak. That’s the point of weighting leading signals. It’s also how you catch the so-called watermelon account: green on the outside, red on the inside.

Turn each band into a playbook

Band	What it means	Playbook	Owner and timing
Green (80 to 100)	Healthy and engaged	Look for expansion, ask for references, share best practices	CSM, at the regular cadence
Yellow (60 to 79)	Watch closely	Identify the weakest component and address it, such as re-engaging users or finding a new champion	CSM, within a week of the change
Red (below 60)	At risk	Escalate, run a recovery plan, involve an executive sponsor, review contract terms	CSM and manager, within 48 hours

A score without a playbook is a dashboard. Write down who does what when an account turns red, and who reviews the list each week. The thresholds and response times here are examples to adapt.

Backtest before you trust it

Many teams build a score and never check it against reality. Before launch, calculate what the score would have been 90 days before each renewal in the last year, then compare it with what happened. Here’s an illustrative result:

Score band 90 days before renewal	Accounts	Churned	Churn rate
Green	120	4	3%
Yellow	50	7	14%
Red	30	15	50%

You want a clear climb in churn from green to red, as above. Two other checks matter. First, calibration: this model flags 15% of accounts as red, close to the overall 13% churn rate. If a score flagged 30% of accounts as at risk while annual churn was in the single digits, the model would be producing noise that your team would learn to ignore. Second, misses: look at the churned accounts that scored green and ask what the model missed. Those are your best clues for improving it.

Building it in HubSpot

For HubSpot teams, a practical setup uses company-level properties and workflows. HubSpot has been adding customer success features, so check what your subscription includes before building anything custom.

• Create component properties on the company record, such as usage trend, adoption breadth, champion status, support score, and sentiment.
• Sync product usage from your analytics tool to the company, not just the contact. Our guide to PLG CRM data unification explains how to match users to accounts.
• Add what HubSpot doesn’t track by default, such as sentiment and champion status, as properties your CSMs fill in or an integration updates.
• Calculate the score with a calculated property or a workflow that combines the components using your weights.
• Trigger action: when an account’s band changes to red, a workflow creates a task for the account owner and notifies the manager.
• Report on it: build a view of accounts by band, renewal date, and ARR, so the biggest at-risk renewals come first.

Dedicated customer success platforms such as Vitally can compute scores from several data sources and write them back to HubSpot. Many teams start with a spreadsheet and CRM properties, then move to a platform once the model has been validated and volumes justify it. AI can help weight signals or flag unusual patterns, but it still needs the backtest and a human who can explain the result. Our piece on AI vs. automation covers where each fits.

Mistakes to avoid

• Averaging whatever data is easy to export, then assigning weights in a meeting without checking whether those signals preceded churn
• Too many inputs, which makes the score impossible to explain or debug
• Leaning on NPS and other lagging signals instead of usage and relationship trends
• Scoring users instead of accounts
• Using one model for every segment
• No playbooks, owners, or response times
• Never recalibrating, even as the product and customers change

Summary

A customer health score is a composite number or color that estimates whether an account will renew, expand, or churn, built from usage, relationship, support, sentiment, and commercial signals. It matters because it turns reactive firefighting into early action and supports net revenue retention. Leading indicators such as usage trends and champion presence deserve more weight than lagging ones like NPS.

To build one, define the outcome, study your own churned and retained accounts, choose five to eight signals, weight them using your data, set thresholds, and attach playbooks with owners and response times. Backtest against past renewals, check calibration, and review the misses. In HubSpot, use company-level properties, synced product usage, and workflows that create tasks for red accounts. Recalibrate every quarter.

Frequently asked

What is a customer health score?+

It’s a composite indicator, such as a 0 to 100 number or a red, yellow, or green label, that estimates how likely an account is to renew, expand, or churn, based on signals like usage, support, relationship, sentiment, and commercial data.

How many signals should a health score use?+

Five to eight is a practical range. More inputs make the score harder to explain and test, and teams stop trusting what they can’t understand.

What weights should I use?+

There’s no universal answer. Published examples often weight usage most heavily, but the right weights come from comparing your churned and retained accounts. Treat any template as a starting hypothesis and test it.

How do I know if my health score works?+

Backtest it. Calculate what each account’s score would have been about 90 days before renewal, then check that churn rises from green to red and that the share of accounts flagged at risk is roughly in line with actual churn.

What is a watermelon account?+

An account that looks green on the surface, perhaps from positive survey scores or steady logins, but is red underneath, such as a champion who left or a quiet drop in key feature use. Weighting leading indicators helps expose them.

Should I score accounts or users?+

Accounts. In B2B SaaS, renewal decisions happen at the organization level, so roll user behavior up to the account and look at patterns like admin activity and champion presence.

How often should we update the model?+

Review it at least quarterly against actual renewals and churn, and whenever your product, pricing, or customer mix changes. Update the underlying data daily or weekly so scores stay current.

Do we need a customer success platform?+

Not at first. Many teams begin with CRM properties and a spreadsheet to validate the model, then adopt a platform once volume and data sources justify it.