Churn almost always looks predictable in hindsight. You look back at a lost client and find three months of declining engagement, a support issue that was never fully resolved, a stakeholder change that nobody tracked, and a renewal conversation that never happened. The warning signs were there. No one was looking at the right combination of signals at the right time.

A customer health score solves this problem by aggregating the signals that matter into a single, actionable indicator. When a client’s score drops, you know to act before they’ve made any decision. When it’s high, you know when to ask for a referral or open an expansion conversation.

What a Customer Health Score Is and What Problem It Solves

A customer health score is a calculated indicator — a number, a color code, or a composite rating — that represents how likely a client is to remain a customer and grow with you. It doesn’t predict the future with certainty, but it reflects the current state of the relationship in a form your team can act on.

The core problem isn’t that account managers don’t care about client health — it’s that health is fragmented across dozens of data points that no one has time to review account by account. A health score compresses that complexity into something your team can scan in a list view and respond to in order of urgency.

Three approaches to health scores:

Weighted numeric score: Assign each health signal a score from 0–100, weight them by importance, and sum to a single number. Green: 70+. Yellow: 40–69. Red: below 40. This approach is precise and allows you to see exactly what’s dragging a score down.

Stoplight model (red/yellow/green): A simpler threshold-based approach where you define specific criteria for each color. Easier to maintain and communicate, less precise for trend analysis.

Composite model: Combines both — a numeric score that rolls up into a stoplight. You see the color at a glance, but can drill into the underlying number to understand why.

For most service businesses, starting with a weighted numeric score that feeds a stoplight display is the right approach.

Choosing the Right Health Signals for Your Business

Not every data point in your CRM is a meaningful health signal. The signals you choose should have a genuine relationship with client outcomes in your specific business. The best way to validate this is to look at your churned accounts and identify which signals were consistently deteriorating before they left.

Engagement Signals

Communication frequency is the baseline signal. If the cadence of calls, emails, and check-ins between your team and the client has dropped significantly below normal without a known reason, that’s a health indicator. Track it as “days since last meaningful contact” rather than total contact volume.

Email response rate and reply lag time reflect the client’s willingness to engage. A client who used to reply within hours and now takes three days is showing a behavioral shift. This doesn’t always mean trouble — they could be busy — but combined with other signals, it becomes meaningful.

Usage and Delivery Signals

Deliverable completion rate measures whether your team is completing its obligations on schedule. If you’re consistently late or incomplete on deliverables, health deteriorates. If deliverables are on track, it’s a stabilizing signal.

Support ticket frequency and topic patterns can reveal systemic problems. A client who opens many tickets about the same area suggests an unresolved underlying issue. Track not just volume but categories of issues.

Relationship Signals

NPS score trend is more informative than a current snapshot. A client who scores 8 three quarters in a row is healthy. A client who has moved from 9 to 7 to 5 over the same period is showing a trend you need to address now, even if 5 isn’t technically a detractor.

Stakeholder changes at the client organization — new decision-makers, reorganizations, or the departure of your primary champion — are among the most disruptive events in a client relationship. Log these in your CRM and weight them heavily in your health calculation for the quarter following the change.

Contract renewal proximity with no renewal conversation yet is a time-based risk signal. As renewal approaches, the absence of a renewal conversation is itself a problem — it means you’re not in control of the timeline.

Assigning Weights and Setting Thresholds

Weights should reflect how strongly each signal correlates with churn in your actual client base. If stakeholder changes have preceded most of your client losses, weight them heavily. If NPS scores haven’t been predictive, weight them lightly.

If you don’t have enough churn history to validate weights, start with equal weights across your signals and adjust over time. An imperfect health score used consistently is better than no health score.

Setting thresholds requires honest conversation with your team about what each level means in practice. Green should mean “this account is stable and the relationship is positive.” Yellow should mean “one or more signals are concerning and proactive action is needed.” Red should mean “this account is at risk and needs immediate intervention.”

Health Signal Reference

Health Signal	Weight (%)	CRM Data Source	Green Threshold	Yellow Threshold	Red Threshold
Days since last meaningful contact	20%	Activity log	Under 21 days	21–45 days	Over 45 days
Email response lag time	10%	Communication log	Under 24 hours	24–72 hours	Over 72 hours
Deliverable completion rate	20%	Task completion fields	90%+ on time	70–89% on time	Below 70% on time
Support ticket volume trend	10%	Support/ticket log	Stable or declining	Moderate increase	Significant increase or repeat issues
NPS score trend	15%	NPS survey field	8+ and stable/rising	6–7 or declining from 8+	Below 6 or rapid decline
Stakeholder change in 90 days	10%	Contact record changes	No change	Change in secondary contact	Change in primary decision-maker
Renewal proximity (days to renewal)	15%	Contract renewal date	Over 90 days	30–90 days, no renewal conversation	Under 30 days, no renewal conversation

Building the Score Inside Your CRM

Using custom fields for health score value: Create a numeric custom field called “Health Score” (0–100) on your account record. This is where the calculated score lives.

Calculated fields vs manually updated scores: Some CRM platforms support calculated fields that automatically update based on other field values. If yours does, use it — an auto-calculating health score removes the dependency on someone remembering to update it. If it doesn’t, you’ll need to set up a manual review workflow where account managers update the score on a defined schedule.

Automating score updates vs manual review triggers: Automated score updates work well for objective, consistently logged signals (last contact date, NPS value). Manual review works better for subjective signals (stakeholder engagement quality, relationship tone). A hybrid approach — automated for objective signals, with a periodic manual review to finalize the score — is pragmatic and accurate.

When to override the automated score: An account manager should always be able to override a health score with a logged explanation. Sometimes a client looks great on paper but the AM knows from a recent call that something is wrong. Override capability preserves human judgment while maintaining data transparency.

How to Act on Each Health Level

Red Accounts

A red health score is a current crisis, not a future risk. The account manager should be notified immediately — not via a dashboard they check weekly, but via an alert that reaches them the same day.

Within 24 hours of going red, a retention call should be scheduled. The call isn’t a sales conversation — it’s a diagnostic one. The goal is to understand what’s wrong before proposing any solution.

If the account manager can’t reach the client within 48 hours, a senior account manager or manager should step in. Escalation isn’t a failure — it’s a deliberate choice to apply more resource to a high-risk situation.

Yellow Accounts

Yellow accounts need proactive outreach within a week, not a month. The goal is to understand the specific signal that’s dragging the score down and address it before it worsens.

A yellow account is a good opportunity for a mid-contract check-in that goes beyond status updates. Ask the client directly: “Is there anything about our work together that you’d want us to do differently?” That conversation, while sometimes uncomfortable, is far better than a surprise cancellation.

Green Accounts

Green doesn’t mean “ignore.” It means this is the right time for deeper investment and forward-looking conversations. Expansion opportunities, referral requests, and case study invitations are all best raised when a client is green.

A client who is in a long green stretch is a candidate for a testimonial or a joint case study. They’re satisfied, they’re not currently navigating a problem with you, and they’re in a frame of mind to talk positively about the relationship.

---

Frequently Asked Questions

How complex should our first health score be?

Start with three to five signals, not eight. A simple system your team actually uses beats a sophisticated system no one updates. Once you’ve been running your health score for two quarters and have learned which signals are most predictive for your business, you can add complexity with evidence to back it up.

How often should the score update?

For automated signals, ideally in real time or daily. For manually reviewed scores, weekly is a reasonable cadence for standard accounts and daily for accounts flagged at-risk. The goal is that your score reflects current reality closely enough that you’re not making decisions based on stale data.

What if we don’t have enough data for weights?

Start with equal weights. Run the system for two quarters. Then compare your health scores at the point of client loss against their subsequent trajectories and identify which signals were most predictive. Use that analysis to adjust weights. You’ll never have perfect data at the start — that’s not a reason to delay building the system.

Should health scores be visible to the client?

Generally no. Health scores are an internal tool for managing your response, not a client-facing metric. Sharing a score implies a judgment about the relationship that can feel clinical or off-putting. The actions you take in response to the score are what the client should experience — proactive outreach, additional attention, improvement conversations — not the score itself.