Health scoring sits in a dangerous place inside Customer Success. Almost everyone agrees it matters, but many teams quietly stop trusting it. The score is too complicated to produce, too dependent on subjective judgment, too disconnected from the way customers actually behave, or too late to be useful. By the time the account turns red, the customer has already made up their mind.
That is why health scoring has to be rebuilt around a different idea. A useful health score is not a label. It is not a color. It is not a dashboard decoration that helps a leadership team feel like someone is watching the portfolio. A useful health score is the visibility layer of the post-sale operating system. It tells the organization whether customers are moving through the experience as designed, whether value is becoming more real over time, and where the system needs to intervene before the renewal conversation becomes a rescue attempt.
If your health score only notices risk when renewal is near, it is not predicting health. It is documenting what already happened.
Most health scores fail because they start at the wrong end of the story. They are built around the visible symptoms of risk, especially the ones that are easiest to measure. Usage drops. Support tickets increase. The CSM feels uneasy. Renewal is approaching. These signals are not worthless, but they are often late. They tell you the customer is already drifting away, not where the drift began.
Churn rarely starts with one dramatic event. It usually begins much earlier, in small moments that compound. Expectations are not realigned after the sale. Onboarding takes longer than the customer expected. The key contact becomes tired or disengaged. First Value is delayed or poorly understood. Usage eventually drops, but by then usage is not the cause. It is the evidence.
This is why health scoring must be connected to the journey, not just the product. Product usage matters, but usage alone does not explain whether the customer is building momentum. A customer can log in often and still fail to reach the outcome that justified the purchase. Another customer may show uneven usage while building executive alignment, completing critical process changes, and preparing for a larger rollout. Without the journey, the score sees activity but misses progress.
The lifecycle plays give health scoring a better foundation. Purchase and Welcome tells us whether the customer was transitioned with confidence. Kickoff tells us whether expectations were reset and the reason for purchase was clarified. Onboarding tells us whether the customer is moving through the early work with enough support and direction. First Value tells us whether the customer has experienced something meaningful quickly enough to believe in the path ahead. Value Blocks, Sharing Insights, Alignment Meetings, and Renew and Grow tell us whether the customer is becoming more capable, more aligned, more visible, and more likely to expand.
A strong health score does not ask, “Is this customer red?” It asks, “Is this customer progressing through the system in the way a healthy customer should?”
This changes the role of health scoring. It is no longer a broad attempt to explain every possible form of churn. It becomes a focused attempt to understand the churn your company can influence. Some churn is unavoidable. A customer may go out of business, be acquired, or outgrow the product because their needs fundamentally changed. Those cases matter to revenue, but they do not create a useful operating signal for the team. The health score should focus on avoidable churn, the kind your company contributes to through slow onboarding, poor communication, product confusion, technical defects, or misalignment between what was sold and what can actually be delivered.
When the score focuses on avoidable churn, it becomes more actionable. It points to the parts of the system that can be improved. It shows whether onboarding is creating fatigue, whether relationship reach is too narrow, whether goal progress is stalling, whether insights are failing to translate into action, or whether the customer still lacks the internal proof needed to defend renewal. The point is not to create a perfect prediction. The point is to create a useful signal early enough to change the outcome.
That requires a different mix of criteria. Traditional criteria still belong in the model. Usage, NPS, support case volume, renewal timing, purchase history, and CSM judgment can all provide context. But they should not carry the whole score. Traditional indicators are often easier to collect than they are to interpret. A usage drop may indicate risk, but it may also indicate seasonality, an implementation dependency, a champion transition, or a process change outside the product. A support spike may indicate frustration, but it may also mean the customer is expanding into more advanced use.
The missing layer is growth-oriented criteria. These are the signals that show whether the customer is becoming healthier before the renewal pressure appears. Relationship reach is one of them. A customer with multiple engaged stakeholders and executive visibility is healthier than a customer dependent on one heroic champion. Lifecycle progression is another. A customer who reaches First Value on time, completes meaningful goal work, and continues to participate in alignment conversations is showing evidence of forward movement. Customer engagement matters as well, especially when the customer consumes guidance, participates in reference activity, shares wins internally, or contributes to a case study. Sophistication of use may be the strongest long-term signal because it shows the customer is not merely using the product but becoming more capable because of it.
The best way to identify these criteria is not to start with the customers who left. That can teach you something, but it often pulls the organization into a failure-centered conversation. A better starting point is your healthiest customers. Look at the customers you would happily clone. They are aligned. They show up prepared. They understand the business objective. They follow through on commitments. They consume guidance. They bring more people into the relationship. They use the product in more mature ways over time. They can describe your value in language their own company understands.
Health scoring should be built by reverse-engineering that pattern. What do healthy customers look like? How do they behave? How do they progress through the lifecycle? How do they engage with your team, your product, your recommendations, and their own internal stakeholders? These questions reveal the criteria that matter because they are grounded in the version of the customer you want to reproduce.
Do not build the score only by studying failure. Build it by understanding what your best customers have in common and designing the system to create more of them.
Once the criteria are identified, the work becomes discipline. Each criterion needs a clear definition. The team needs to know what it means, why it matters, how it will be measured, and what a low or high score actually represents. This is where many health scores become vague. They rely on words like engagement, alignment, or value without defining what those words mean in the operating system. If two CSMs would score the same customer differently because the criteria are unclear, the model is not ready.
Weighting is where the model becomes more honest. Not every signal has equal predictive power. A weak champion may matter more than a modest support increase. Delayed First Value may matter more than a temporary usage dip. Lack of executive visibility may matter more than a neutral NPS response. The weight should reflect the degree to which each criterion helps predict future behavior and the degree to which the company can act on it. The score does not need to be mathematically perfect, but it does need to be thoughtful enough to earn trust.
That trust is built through validation. Score a representative set of customers, then look at the extremes. Are the strongest green accounts truly healthy? Are the deepest red accounts truly at risk? Where does the score surprise the team in a useful way? Where does it misrepresent reality? This is not a failure of the model. It is how the model learns. Health scoring is not something you finish. It is something you launch, use, inspect, and refine.
This is also why a health score must come from daily operations. If the score depends on a special project, a quarterly cleanup, or a manual scramble before an executive meeting, it will disappear. The mechanisms for capturing the score have to be embedded in the work itself. The kickoff should create data. Onboarding should create data. First Value should create data. Goal progress should create data. Alignment meetings should create data. The more health scoring becomes a byproduct of the system, the more trustworthy it becomes.
A simple, live health score that inspires action will always beat a sophisticated score that no one uses. Simplicity does not mean shallow. It means focused. The model should contain enough criteria to tell a useful story, but not so many that the team loses the plot. It should be current enough to support decisions, but not so reactive that every small movement creates noise. Above all, it should point to action. If the score does not help the team decide what to do next, it is not an operating tool.
A health score that does not change behavior is just reporting. A health score that works tells the team where to act and why it matters.
At this point, the challenge becomes practical. How do you take the signals that matter, define them clearly, weight them thoughtfully, and test whether they reflect reality? How do you avoid building another spreadsheet that becomes too hard to maintain? How do you move from concept to prototype without waiting for perfect data?
To make that transition easier, we built the Health Score Builder for this system. It is designed to help leaders move from abstract agreement to a working model. The tool helps you select a focused set of criteria, balance traditional and growth-oriented signals, assign weights, test customer examples, and begin the rollout process with more confidence.
The tool is not the health score by itself. It is the translation layer that helps the operating system become visible. It works because the journey is already defined, the plays already describe the intended customer experience, and the team has a clearer view of what healthy progress should look like. Without that foundation, a score is just a number. With it, the score becomes a signal the business can act on.
This is the real purpose of health scoring. Not to prove that Customer Success is watching the customer base. Not to make a dashboard look more complete. Not to create a false sense of precision. The purpose is to help the company see customer progress early enough to protect it, accelerate it, and scale it.
A health score that actually works does not merely tell you who is red. It tells you where the system needs to act.