Customer Health Scores: What They Measure and What They Miss

A customer health score is a composite metric that tells you how likely a customer is to stay, expand, or churn. It aggregates signals like product usage, support ticket volume, NPS responses, and contract engagement into a single number that gives revenue and marketing teams a shared view of account risk. Done well, it is one of the most commercially useful tools in a growth operator’s kit. Done badly, it becomes a dashboard metric that everyone monitors and nobody acts on.

I have spent a good portion of my career inside businesses where the relationship between customer satisfaction and commercial performance was either poorly understood or actively ignored. Marketing was being used to fill a leaky bucket, running acquisition campaigns that brought in customers who were gone within six months. Nobody was measuring why. Nobody had a health score. Nobody had a systematic view of what a good customer looked like versus one who was about to leave. The growth strategy conversation has to start with retention before it earns the right to talk about acquisition, and that framing sits at the heart of how I approach the Go-To-Market and Growth Strategy content on this site.

What Does a Customer Health Score Actually Measure?

A health score is a weighted model. You select a set of signals that correlate with retention or churn, assign each one a weight based on its predictive importance, and combine them into a single score, usually expressed on a 0-100 scale or a traffic light system. The signals themselves fall into a few broad categories.

Product engagement is the most common input. How frequently is the customer logging in? Which features are they using? Are they using the features that your most successful customers rely on, or are they stuck in a shallow part of the product? Usage depth tends to be more predictive than usage frequency. A customer who logs in every day but only uses one basic function is not necessarily healthy.

Relationship signals cover the human side of the account. Are they responsive to their customer success manager? Have they attended your QBR? Did they renew on time last cycle or did it drag? These signals are harder to systematise but often highly predictive. An account that goes quiet three months before renewal is rarely a good sign.

Support data tells you about friction. High ticket volume can indicate an engaged customer who is pushing the product hard, or it can indicate a customer who is struggling and losing patience. Context matters. A spike in tickets after onboarding is different from a spike eight months into a contract.

Financial signals round out the picture. Are they paying on time? Have they expanded their contract? Are they using the volume they purchased, or are they significantly under their committed usage? Significant under-utilisation is one of the cleaner leading indicators of churn, because it tells you the customer has not integrated your product into their workflow at the depth required to justify renewal.

Why Most Health Scores Fail Before They Are Even Built

The failure mode I see most often is not technical. It is strategic. Teams build health scores around the data they have rather than the signals that matter. They use login frequency because it is easy to pull from the product database. They use NPS because they already run the survey. They exclude signals that would require cross-functional data sharing because that conversation is too difficult to have.

The result is a score that looks credible but has no real predictive validity. It correlates with what you already know rather than telling you something new. I have sat in enough strategy reviews to recognise this pattern quickly. The score goes green, the account churns, and the post-mortem reveals that the signals everyone knew mattered, the ones that lived in the CRM or the finance system or the support platform, were never included.

There is also a subtler problem. Many health scores measure activity when they should measure outcomes. A customer who is highly active in your product is not necessarily getting value from it. Value realisation is the outcome that drives retention. If your health score cannot distinguish between a customer who is active and getting results versus one who is active and frustrated, it is not measuring health. It is measuring busyness.

This connects to a broader point about how we use data in marketing and commercial strategy. Behavioural data from tools like Hotjar can tell you what customers are doing, but interpreting what that behaviour means requires judgment and context. The data is a perspective on reality, not reality itself. Health scores inherit that limitation.

How to Build a Health Score That Has Predictive Value

Start with churn analysis, not with signals. Look at the customers who churned in the last 12 to 24 months and work backwards. What did they have in common? What signals, had you been tracking them, would have indicated risk three months before the decision was made? That exercise tells you which signals actually matter for your specific business, rather than which ones appear in generic health score templates.

Then do the same exercise with your best customers. What does a healthy account look like at 90 days? At 12 months? What features are they using? How are they engaging with your team? What does their support history look like? The contrast between your best accounts and your churned accounts gives you the signal architecture for a score that is grounded in your actual commercial reality.

Weight the signals based on their predictive importance, not their availability. This requires some honest internal negotiation. The signals that matter most are often the ones that require cross-functional data sharing, which means getting product, finance, and customer success aligned on a shared data model. That is a harder conversation than pulling a login report, but it is the conversation that makes the score useful.

Set thresholds that mean something. A score of 65 should trigger a specific action. A score of 40 should trigger a different, more urgent action. If your thresholds do not map to a response protocol, you have a reporting tool, not a management tool. The score is only valuable if it changes what someone does on Tuesday morning.

Validate the model against actual outcomes. Once you have been running the score for a quarter or two, check whether the scores you assigned six months ago predicted what actually happened. Did the accounts you scored red actually churn at a higher rate? Did the accounts you scored green actually renew and expand? If the correlation is weak, the model needs revision. This is not a sign of failure. It is the normal process of building something that works.

Where Marketing Fits Into the Health Score Framework

Marketing’s role in customer health is underappreciated and often poorly defined. In most organisations, marketing hands off to sales at the contract stage and considers its job done. Customer health becomes the domain of customer success, and marketing re-enters the picture only when there is an upsell or renewal campaign to run. That is too narrow a view.

Onboarding is a marketing problem as much as a product or customer success problem. The communications, content, and sequencing that take a new customer from signed contract to active user are marketing decisions. If customers are not activating properly, the health score will reflect that within 60 days. Marketing should own the onboarding communication layer and optimise it with the same rigour applied to acquisition campaigns.

At-risk re-engagement is another area where marketing has a direct role. When a health score drops below a threshold, the response is typically a customer success call. That is right. But it should also trigger a content and communication response, relevant case studies, feature adoption guides, community invitations, whatever helps the customer reconnect with the value they signed up for. Marketing should have pre-built plays for each health score segment, not just for each funnel stage.

Expansion signals are worth calling out separately. Customers who are heavily using one part of your product but have not adopted adjacent features are a natural expansion audience. Their health score data tells you exactly what they are doing and what they are not. Marketing can use that to run highly targeted campaigns that feel genuinely helpful rather than generic upsell pressure. That is the kind of personalisation that actually earns the name.

I have seen this done well at companies that treat their customer base as a growth asset rather than a maintenance burden. The difference in commercial performance is significant. BCG’s work on go-to-market strategy makes the point that the intersection of marketing and customer experience is where the most durable growth happens. Health scores are one of the mechanisms that make that intersection operational rather than theoretical.

The Organisational Dynamics That Determine Whether Health Scores Work

A health score is a cross-functional instrument. It draws data from product, finance, support, and customer success. It should inform decisions made by marketing, sales, and leadership. That means it requires genuine cross-functional ownership, which is where most implementations run into difficulty.

In my experience running agencies and working with clients across more than 30 industries, the single biggest predictor of whether a health score programme delivers commercial value is whether there is a named owner who has the authority and the relationships to pull data from multiple systems and drive action across multiple teams. Without that, the score becomes a dashboard that customer success watches and everyone else ignores.

The second organisational factor is response protocol clarity. Every score threshold needs a defined owner and a defined action. Red accounts need a specific escalation path. Yellow accounts need a specific intervention cadence. Green accounts with expansion signals need a specific outreach approach. If the protocol is vague, the score is vague. People will look at it, nod, and go back to whatever they were doing before.

There is also a cultural dimension that rarely gets discussed. Health scores can create perverse incentives if they are tied too tightly to individual performance metrics. A customer success manager who knows their accounts are being scored may start managing the score rather than managing the relationship. That means coaching customers on NPS responses, pushing feature adoption for its own sake, or avoiding difficult conversations that might generate support tickets. The score should inform management decisions, not replace management judgment.

I watched a version of this play out at a client we worked with on their go-to-market strategy. They had built a reasonably sophisticated health score, but the customer success team had learned which inputs drove the score and were optimising for those inputs rather than for genuine customer outcomes. The score looked healthy. The renewal rate told a different story. The fix was partly technical, adding signals that were harder to game, and partly cultural, changing how the score was used in performance conversations.

Health Scores Across Different Business Models

Health scores originated in SaaS, where the subscription model creates a natural need for churn prediction. But the underlying logic applies more broadly than the SaaS community sometimes acknowledges.

In professional services and agency businesses, client health can be tracked through a similar framework. Engagement frequency, scope expansion, payment behaviour, referral activity, and the quality of briefing conversations are all signals. I ran agencies for years and wish I had formalised this earlier. We had intuitive views on which clients were at risk, but no systematic model. The ones that churned were rarely a surprise in hindsight. They were often a surprise in practice because nobody had aggregated the signals into a clear picture.

In e-commerce and retail, health scores translate into customer lifetime value models and engagement scoring. Purchase frequency, category breadth, return rates, email engagement, and loyalty programme participation are all inputs. The mechanics are different from a SaaS health score, but the commercial logic is identical. You are trying to identify which customers are at risk of lapsing before they lapse, so you can do something about it.

For B2B companies with longer sales cycles and complex accounts, health scores become particularly important because the cost of losing a major account is disproportionately high. Forrester’s analysis of go-to-market challenges in complex industries points to the difficulty of maintaining commercial momentum across long relationship cycles. A health score is one of the tools that keeps that momentum visible and manageable.

The Honest Limitations of Customer Health Scores

No model captures everything. Health scores are built on the data you have access to, weighted by assumptions you have made about what drives retention in your business. Those assumptions may be wrong, and the data may be incomplete. A customer can score green and churn because their internal champion left the company, because their budget was cut, because a competitor made them an offer that was commercially impossible to refuse. None of those signals live in your product database.

This is not an argument against health scores. It is an argument for building them with appropriate humility. The score is a probability estimate, not a prediction. It raises or lowers the likelihood of a particular outcome. It does not determine it. Teams that treat a green score as a guarantee of renewal will be caught out. Teams that treat a red score as a definitive churn signal will waste resources on accounts that were never actually at risk.

There is also a data quality problem that compounds over time. If your product telemetry is unreliable, if your CRM data is inconsistently maintained, if your NPS response rates are low, the score will reflect those gaps. A health score is only as good as the data discipline behind it. That sounds obvious, but the number of organisations running sophisticated-looking health score dashboards on top of patchy, inconsistently maintained data is higher than anyone in a vendor presentation will admit.

Growth strategy thinking, whether you are drawing on frameworks from growth-focused practitioners or building your own commercial model, tends to be optimistic about what measurement can tell you. The honest position is that measurement gives you better approximations, not certainties. Build your health score accordingly, and use it as one input into commercial judgment rather than a replacement for it.

If you are thinking about where health scores fit within a broader commercial framework, the Go-To-Market and Growth Strategy hub covers the adjacent decisions around retention, expansion, and how marketing earns its place in the revenue conversation.

Getting Started Without Overcomplicating It

The temptation when building a health score for the first time is to make it comprehensive. Every signal, every data source, a sophisticated weighting algorithm. Resist that temptation. Start with three to five signals that you are confident are meaningful, build a simple model, and run it alongside your existing processes for one quarter. See whether it tells you anything you did not already know. See whether it changes any decisions. Then iterate.

A simple model that gets used is worth more than a sophisticated model that sits in a dashboard and gets checked once a month. The commercial value of a health score is entirely downstream of the actions it triggers. If it is not triggering actions, it is not delivering value, regardless of how technically impressive it is.

The other thing I would say to anyone starting this process is to involve your customer success team from the beginning, not as data providers but as co-designers. They carry a large amount of qualitative knowledge about what healthy and unhealthy accounts look like. That knowledge is difficult to quantify but essential context for building a model that reflects reality. The best health score implementations I have seen were built through genuine collaboration between data, product, and customer-facing teams. The worst were built by a data team working in isolation and handed over as a finished product.

Customer health scoring is not a silver bullet. But for businesses where retention and expansion drive a meaningful share of revenue growth, it is one of the more commercially grounded tools available. Build it carefully, use it honestly, and treat it as a living model rather than a finished one.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions