Customer Engagement Score: What It Measures and What It Misses

A customer engagement score is a composite metric that quantifies how actively a customer interacts with your product, content, or brand over a defined period. It typically combines signals like login frequency, feature usage, email opens, support interactions, and purchase behaviour into a single number that represents relationship health. Done well, it gives revenue and marketing teams an early warning system for churn and an objective basis for prioritising accounts.

Done poorly, it becomes a number that looks precise and means very little. That gap between the two is where most companies actually live.

Why Engagement Scores Exist in the First Place

When I was running agencies, the closest equivalent we had to a customer engagement score was a gut read from account directors. They knew which clients were engaged, which were drifting, and which were about to call a review. The problem was that gut reads don’t scale, they don’t transfer when someone leaves the business, and they’re invisible to anyone in finance or operations who needs to plan ahead.

Engagement scores exist because businesses need a systematic, repeatable way to answer one question: is this customer getting value from us, and are they likely to stay? That question matters more as a business grows. When you have 20 clients, you know them. When you have 2,000, you need a model.

The logic is straightforward. Customers who engage regularly with a product or service tend to renew, expand, and refer. Customers who go quiet tend to churn. If you can spot the early signals of disengagement, you can intervene before the cancellation email arrives. That’s the commercial case for building the score in the first place.

This sits squarely within the broader discipline of go-to-market execution. If your go-to-market and growth strategy doesn’t include a clear view of how you’ll retain and grow existing customers, you’re building a leaky bucket and calling it a pipeline.

What Goes Into a Customer Engagement Score

There’s no universal formula. The signals that matter depend entirely on your product, your customer type, and what value actually looks like for your business. That said, most engagement score models draw from a similar set of input categories.

Product or platform activity is typically the strongest signal in SaaS and digital-first businesses. Login frequency, session depth, feature adoption, and time-in-product all indicate whether the customer is genuinely using what they’re paying for. A customer who logs in daily and uses five features is almost certainly in a different position than one who hasn’t logged in for three weeks.

Communication and content engagement covers email open and click rates, webinar attendance, documentation reads, and response rates to outreach. These are weaker signals on their own but useful in context. Someone who reads every product update email and attends your user conference is probably invested in the relationship.

Commercial signals include payment history, contract renewal timing, expansion purchases, and the number of seats or licences in use relative to what’s been purchased. A customer who has used 80% of their licence allocation is a different commercial prospect than one at 20%.

Support and service interactions are more nuanced. High support volume could mean the customer is deeply engaged and running into problems. It could also mean the product isn’t working for them. Context matters here more than the raw number.

Relationship signals include NPS responses, participation in advisory boards, case study requests, and referrals. These are high-quality signals but low-frequency, so they tend to act as multipliers rather than base inputs.

How to Weight the Signals Without Fooling Yourself

Weighting is where most engagement score projects go wrong. The temptation is to assign weights based on intuition, or worse, based on what data is easiest to pull. Neither produces a score that actually predicts commercial outcomes.

The right approach is to start from the outcome you care about, typically retention or expansion revenue, and work backwards. Look at your churned customers over the past 12 to 24 months. What signals dropped before they left? Look at your best-retained, highest-expansion accounts. What behaviours did they share? The weights should reflect what actually predicts the outcome, not what feels intuitively important.

I’ve seen this done badly in a few different ways. One common version is over-weighting email engagement because it’s easy to track. A customer who opens every email but never logs into the product is not engaged in any commercially meaningful sense. Another version is weighting all signals equally because it feels fair. Equal weighting is usually wrong. Not all signals carry the same predictive value.

The most defensible approach is to build an initial model based on your best hypothesis, validate it against historical data, then adjust. Treat it as a working model, not a finished product. Forrester’s research on intelligent growth models makes a similar point: the companies that grow sustainably are those that build feedback loops into their measurement frameworks, not those that set a model and forget it.

The Segmentation Problem Most Teams Ignore

A single engagement score applied uniformly across your entire customer base is almost always misleading. A mid-market customer using your product for internal reporting has different usage patterns than an enterprise customer running it across 12 departments. Scoring them on the same scale produces numbers that aren’t comparable and interventions that don’t fit.

The fix is segmentation. Build separate scoring models, or at minimum separate benchmarks, for distinct customer segments. The obvious cuts are by company size, industry, and product tier. But lifecycle stage matters just as much. A customer in their first 90 days should be scored differently than one who has renewed three times. Early-stage engagement is about onboarding completion and initial value realisation. Later-stage engagement is about depth of use and expansion potential.

When I was managing large-scale performance marketing programmes across multiple industries, one of the consistent lessons was that aggregate metrics hide more than they reveal. A campaign that’s performing at average across 10 segments might be excellent in three and failing in seven. The average obscures both the success and the problem. Engagement scores work the same way.

BCG’s work on go-to-market strategy points to a similar principle: growth strategies that treat customers as a homogeneous group tend to underperform those that differentiate by segment and lifecycle position. Engagement scoring is a tactical application of that same logic.

What a Good Engagement Score Actually Enables

A well-built engagement score isn’t just a reporting metric. It’s an operational tool. The value comes from what it allows teams to do differently.

For customer success teams, it creates a prioritisation framework. Instead of working through accounts alphabetically or by revenue size alone, CSMs can focus attention on accounts where the score is declining or where it’s never reached a healthy baseline. That’s a much more defensible way to allocate a finite resource.

For sales teams, high engagement scores in existing accounts are a signal for expansion conversations. A customer who is deeply embedded in your product, using multiple features and expanding user counts, is a far better expansion prospect than one who is barely using what they’ve already bought. Vidyard’s research on untapped pipeline potential for GTM teams points to exactly this: a significant share of expansion revenue is sitting inside existing accounts that aren’t being worked systematically.

For marketing, engagement scores can inform campaign targeting. Customers in a healthy score range are candidates for advocacy programmes, referral campaigns, or upsell sequences. Customers with declining scores need a different kind of communication, one focused on re-establishing value rather than selling more.

For leadership, aggregate score trends across the customer base give an early read on retention risk before it shows up in churn numbers. If average engagement scores are declining across a cohort, that’s a leading indicator worth acting on before it becomes a lagging one.

The Limits of the Score and Where It Breaks Down

I want to be honest about what engagement scores can’t do, because the category has attracted a certain amount of vendor hype that doesn’t serve anyone well.

An engagement score measures behaviour, not sentiment. A customer can be highly active in your product and still be looking for alternatives. They might be deeply embedded because switching costs are high, not because they’re satisfied. Engagement is a proxy for value, but it’s an imperfect one. You still need qualitative data, actual conversations, NPS follow-ups, and business reviews to understand the full picture.

Scores also don’t capture relationship quality at the human level. In B2B, a lot of retention comes down to whether the customer’s key contact trusts your team. That doesn’t show up in a login frequency metric. When I was at Cybercom, early in my career, there was a Guinness brainstorm where the founder had to leave for a client meeting and handed me the whiteboard pen. The internal reaction in the room was quiet panic. But the point is that the client relationship with the founder was built on something entirely unmeasurable. No engagement score would have captured what that relationship was worth.

There’s also a gaming risk. Once teams know what drives the score, behaviour can shift to optimise the score rather than the underlying outcome. If login frequency is heavily weighted, you might see customers encouraged to log in without actually doing anything meaningful. That’s a measurement problem, not a customer health problem, and conflating the two is dangerous.

The broader point is one I’ve made before: analytics tools give you a perspective on reality, not reality itself. A customer engagement score is a model. It has assumptions baked in. Those assumptions should be tested regularly against what actually happens to customers over time.

Building the Score Without Overcomplicating It

Most teams that build engagement scores for the first time make them too complicated. They identify 20 signals, assign precise decimal weightings, and produce a score that nobody in the business understands or trusts. Complexity is not sophistication.

Start with five to seven signals that you’re confident are meaningful. Weight them based on your best hypothesis about what predicts retention. Produce a score on a simple 0 to 100 scale. Define three or four health bands, something like at-risk, needs attention, healthy, and thriving, and make sure the bands map to specific actions for the teams using the score. Then run it for a quarter and check whether the scores are actually predicting what you expected them to predict.

The goal in the first iteration is not to build the perfect model. It’s to build a model that’s better than gut feel and simple enough that people will actually use it. You can refine from there.

One thing that consistently helps is involving the commercial team in the design process from the start. Customer success managers know which customer behaviours actually matter. Sales knows which accounts expand and why. If you build the score in a data team silo and then present it to the commercial team as a finished product, you’ll get polite scepticism at best and active resistance at worst. The teams who will use the score should have a hand in shaping it.

Growth hacking frameworks, like those documented at Crazy Egg, often treat engagement metrics as a core loop: measure, act, measure again. That rhythm applies directly to engagement scoring. The score is not a destination. It’s a feedback mechanism.

Connecting Engagement Scores to Revenue Outcomes

The reason engagement scores matter commercially is that they sit upstream of revenue events: renewals, expansions, and referrals. If you can shift the distribution of your customer base toward higher engagement scores, you should see downstream improvement in net revenue retention.

That connection needs to be made explicit in how you report the score internally. A score that lives only in a customer success dashboard is a CS metric. A score that’s tied to renewal forecasts, expansion pipeline, and churn risk projections becomes a commercial metric. That shift in framing changes how seriously leadership takes it and how much resource gets allocated to improving it.

One practical way to make the connection is to run a cohort analysis. Take customers who were in your top engagement score quartile 12 months ago and look at their renewal and expansion rates today. Do the same for the bottom quartile. If the score is working, you should see a meaningful difference. If you don’t, the model needs revision.

BCG’s perspective on scaling agile practices is relevant here in an adjacent way: the organisations that get the most from data-driven tools are those that build tight feedback loops between measurement and action. An engagement score without a defined response playbook is just a number. The playbook is what turns it into a growth mechanism.

If you’re thinking about where engagement scoring fits within a broader commercial strategy, it belongs in the same conversation as your retention programmes, your expansion motion, and your customer marketing approach. These aren’t separate workstreams. They’re different expressions of the same underlying question: are we delivering enough value to keep customers and grow them? The go-to-market and growth strategy thinking on this site covers that broader frame in more detail.

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