Your Best Retention Metric Isn't a Loyalty Score
Welcome back to this week’s newsletter.
This week, I want to explore a common mistake many retention teams make: assuming loyalty scores can predict churn. In reality, they don’t, at least not in the way most people think.
You survey your customers, ask if they’d recommend you, see promising scores, and then observe those same customers leave. The data isn’t wrong; you’re simply asking a question that doesn’t predict the outcome you’re trying to prevent.
Loyalty scores measure intent, but customer retention depends on predicting actions. A customer who gives a high rating and then cancels the next month isn’t being dishonest; they truly meant it at the time.
However, their budget might have been cut, a competitor could have offered a better deal, or they realized they weren’t making full use of what they paid for. Their intent remains the same, but changing circumstances lead to cancellations.
The Three Things You’re Conflating
When you ask “Are our customers loyal?” you’re actually asking three separate questions:
Will they stay?
Will they recommend us?
Will they spend more?
These responses rarely align as you might expect. Someone can love your product and still leave, complain constantly but never actually churn, or feel completely neutral while quietly increasing their contract size.
The person providing glowing feedback and the one about to renew for another year are often different, and treating them as a single metric can dilute the essential signals you need.
Loyalty retention focuses on preventing customers from leaving. Advocacy loyalty involves encouraging customers to recommend your brand rather than remain silent. Purchasing loyalty aims to increase the amount customers buy rather than retain them.
Each type reacts to different triggers, follows different timelines, and needs different strategies. Measuring one without considering the others can lead to surprises in churn rates, even when you believed you had prevented them.
What Predicts Retention Better Than Asking About Retention
Behavior patterns beat stated intent almost every time.
Login frequency has declined over the past 90 days, and feature adoption has stalled since onboarding. Support tickets spike and then go quiet. Usage is focused on one user despite the account having 20 seats.
A champion who used to respond within hours now takes days. These indicators provide more insight into churn risk than simply asking users to rate their likelihood of switching.
By the time a customer consciously knows they’re going to leave, the decision is already made, and the behavioral signals showed up months earlier. They just weren’t being tracked, or they were sitting in a dashboard nobody thought to connect to retention outcomes.
The gap between what customers say and what they do isn’t dishonesty, it’s the difference between how people feel in a moment versus how they act over time. Surveys capture a snapshot while behavior tells the story.
Where Loyalty Scores Still Matter
This isn’t an argument to stop measuring NPS or satisfaction; it’s an argument to stop expecting them to do a job they weren’t designed for.
Advocacy scores help you understand referral potential: who might actively refer new business to you versus who’s satisfied enough to stay quiet but won’t lift a finger to promote you. That’s valuable for understanding organic growth potential and identifying customers worth featuring in case studies or referral programs.
Expansion scores help identify when to upsell by showing who is ready to grow versus who is still trying to prove value internally. Asking about purchasing intent can reveal opportunities your CS team might otherwise miss.
But for retention specifically, it’s better to observe what customers actually do rather than asking them what they’ll do, because behavioral signals are more honest, timely, and less likely for customers to misrepresent, even if they wanted to.
What Your Early Warning System Actually Needs
Your quarterly survey isn’t built to be a retention forecasting tool, so stop asking it to be one.
Build your early warning system from usage data, engagement patterns, and the behavioral signals that precede churn - the things customers can’t misreport because they’re not reporting them at all.
Track what they’re doing, or more importantly, what they’ve stopped doing.
Use loyalty scores for what they’re truly effective at: advocacy scores indicate referral and reputation potential, expansion scores show upsell readiness, and retention scores (if you’re asking for them) should be checked against actual churn data to see if they are even predictive for your specific customer base.
Different questions require different data sources, and the mistake is mixing them into a single metric while expecting that metric to tell you everything.
A system that separates intent from behavior and matches each metric to the outcome it actually predicts gets you much closer to seeing what’s coming before it arrives.