A customer health score combines several signals into a single indicator of whether an account is likely to renew, expand or churn. Done well, it gives customer success and account management early warning, so they act months before a renewal rather than days. Done badly, it becomes a colourful field nobody trusts. The difference lies in choosing signals that actually move before customers leave, weighting them from your own history, and connecting each score band to a specific action. This article walks through those steps and shows how to build the score in HubSpot and Zoho CRM. It goes deeper than the overview in the churn reduction playbook.
Step 1: list candidate signals
Start with every signal you can observe that might relate to retention. Usage: logins, active users against seats purchased, use of key features, volume of core actions. Breadth: number of teams or departments using the product, number of integrations connected. Relationships: whether the champion and executive sponsor are still in role and active, attendance at business reviews. Support: ticket volume and severity, time to resolution, escalations. Sentiment: survey scores, feedback in calls. Commercial: payment delays, requests for discounts, downgrades. Outcomes: whether the customer is achieving the goal they bought for, if you can measure it. Not all of these will be available or useful, but listing them shows what data you need to bring into the CRM.
Step 2: test signals against churned accounts
The signals that matter are those that change before churn, not after. Pull a list of customers who churned or downgraded over the last year or two, and a comparison group who renewed. For each, look at the candidate signals around three to six months before the renewal date. Which signals looked different for the churned group? Common findings include falling usage, loss of the champion, unresolved high-severity tickets and missed business reviews. Keep the three to six signals that best separate the groups. If you have few churned customers, combine this analysis with the judgement of customer success managers, and refine it as more data accumulates.
Step 3: weight the signals and set thresholds
Give each signal points or a weight based on how strongly it separated churned and renewed accounts. Keep the model simple enough to explain: a CSM should be able to see why an account is at risk. Convert each signal into a score using clear rules, for example usage above a set share of purchased seats earns full points, below a lower threshold earns none. Sum the scores and set bands for healthy, neutral and at risk. Check the bands against your historical accounts: most churned accounts should have been at risk well before they left, and most healthy accounts should have renewed. Adjust until the score would have warned you early on most past churn.
Step 4: connect each band to a play
A score without an action is just a number. Define what happens at each band.
| Band | What it means | Play | Owner |
|---|---|---|---|
| Healthy | Signals on track | Look for expansion, ask for referral or case study | Account manager |
| Neutral | One or two signals slipping | Check-in call, targeted training, confirm goals | CSM |
| At risk | Several signals weak or a critical one failed | Recovery plan, executive outreach, review value delivered | CSM and manager |
| Critical event | Champion left, major escalation, payment issue | Immediate review regardless of score | CS lead |
Building it in HubSpot
HubSpot's customer success workspace, part of Service Hub, includes health scores built from property values and tracked events, with scoring groups, point caps and thresholds for health categories. According to HubSpot's knowledge base, the health score card on records needs Service Hub Professional or Enterprise and a Service seat. To use it, bring the signals into HubSpot first: product usage as company or contact properties synced from your product or data warehouse, ticket data from the help desk, survey responses and engagement from meetings. Then configure the score with the weights and thresholds from step 3, and use workflows to create tasks or alerts when an account moves into a lower band. If you do not have the required edition, a calculated property and workflows can approximate the same logic.
Building it in Zoho CRM
In Zoho CRM, health scores are usually built with a combination of custom fields, formula fields or scoring rules, depending on your edition, and workflow rules. Sync usage data from your product into fields on the account record, using Zoho's APIs, Zoho Flow or a data integration tool. Bring ticket data in from Zoho Desk or your help desk. Calculate each signal's points in a formula field or scoring rule, add them into a total health score, and use a further formula or workflow to set the band. Workflow rules can then assign tasks, send alerts to the account owner or update a renewal risk field. Keep the logic documented outside the CRM as well, because formula fields can be hard to read later.
Getting usage data into the CRM
The hardest part of most health scores is product usage data, which lives in the product database or analytics tool rather than the CRM. Options include a scheduled sync from your data warehouse to the CRM using a reverse ETL tool, a direct integration from your product analytics tool, or a simple scheduled export and import for early stages. Sync aggregated metrics, such as active users in the last 30 days or key actions per week, rather than raw events, to keep the CRM fast and the logic clear. Agree with engineering how often data updates and who owns the sync, because a health score built on stale usage data gives false comfort.
Reviewing and improving the score
Review the score each quarter. Compare it with actual renewals and churn: did at-risk accounts churn more often than healthy ones, and how early did the score warn? Ask CSMs where the score disagreed with their judgement and why. Adjust signals, weights and thresholds accordingly, and record the changes so trends over time remain interpretable. Also review the plays: are at-risk accounts that received a recovery plan renewing more often? The goal is a score that customer success trusts enough to plan their week around.
Health scores for smaller customer bases
Companies with a small number of large customers sometimes assume a health score is unnecessary, because the team knows every account. Knowing an account is not the same as seeing early warning signs consistently, especially as the team grows or account owners change. With few customers, keep the score simpler: three or four signals, assessed partly by the account owner each month, combined with any usage data available. A short monthly health review, where each owner rates their accounts against the same criteria and notes changes, can work as well as an automated score at this size, and it builds the habit and the history that a more automated model will need later. Record the ratings in the CRM rather than a separate document, so the history is visible when someone new takes over the account.
FAQ
A combined indicator built from signals such as usage, adoption, relationships, support and sentiment, used to predict whether an account will renew, expand or churn.
Usually three to six. Choose those that changed months before churn in your own history. More signals make the score harder to explain and maintain.
Yes, in the customer success workspace in Service Hub. HubSpot's knowledge base says the health score card needs Service Hub Professional or Enterprise and a Service seat.
Yes, with custom and formula fields or scoring rules depending on edition, usage data synced into account records, and workflow rules for alerts and tasks.
Review it quarterly against actual renewals and churn, and adjust signals, weights and thresholds when the score failed to warn early.
Product usage, because it lives outside the CRM. Sync aggregated usage metrics on a schedule rather than raw events.
Generally not the score itself, but sharing the underlying measures in business reviews, such as adoption against goals, helps customers act on the same signals your team watches.
Customer success owns the model and the plays, with RevOps maintaining the data feeds, fields and automation behind it. Agree who approves changes to signals and weights so the score stays consistent.