Why Points Don't Measure Loyalty (And What Does)

In 2009, a Delta SkyMiles member with a six-figure mileage balance and Diamond Medallion status switched his corporate travel account to a competitor within a single quarter — the moment a rival airline offered a marginally more convenient route to his new client's city. He had every incentive the points system could offer: a mile balance worth thousands of dollars, lounge access, upgrade priority. None of it held him. He wasn't loyal. He was stuck, and the moment the cost of staying exceeded the cost of leaving, he left.

That is the quiet scandal at the heart of most loyalty measurement: programmes report points issued, points redeemed, tier migration and member counts, and call the resulting dashboard "loyalty." It isn't. It's a ledger of transactions. Loyalty is measured by what a customer does when the deal is no longer the best deal — not by how many points they're sitting on. If you want to know whether you have real loyalty, you have to measure attachment, advocacy and resistance to switching, not just redemption rates.

Why don't points balances measure loyalty?

Points measure behaviour that has already happened, usually behaviour that was bought. A customer scanning a loyalty card at the till is exhibiting what retention economists call mercenary loyalty — repeat purchase driven entirely by the next incentive, with zero emotional residue once the incentive disappears. The points programme is doing exactly what it was built to do: it is just not building what most executives assume it's building.

Two behavioural mechanisms explain why points can inflate apparent loyalty while actual attachment stays flat or erodes:

• Loss aversion keeps members transacting, not bonding. Daniel Kahneman and Amos Tversky's prospect theory, published in their 1979 paper "Prospect Theory: An Analysis of Decision under Risk" in Econometrica, established that people feel the pain of a loss roughly twice as intensely as the pleasure of an equivalent gain. A member sitting on 40,000 unredeemed points isn't emotionally invested in the brand — they're avoiding the felt loss of forfeiting a balance. That's a very different thing from wanting to come back.
• The goal-gradient effect accelerates effort near a reward, then evaporates. Ran Kivetz, Oleg Urminsky and Yuhuang Zheng's 2006 study "The Goal-Gradient Hypothesis Resurrected," published in the Journal of Marketing Research, found that customers holding a loyalty card towards a free car wash washed their cars measurably more often as they approached the reward threshold — then slowed down immediately after redeeming. The card was driving a sprint to the finish line, not a relationship.

Neither mechanism is a flaw in the customer. They're predictable responses to a mechanic built on redemption thresholds. The problem is that most loyalty dashboards were designed to track the mechanic, not the mindset behind it.

What's the real difference between behavioural loyalty and emotional loyalty?

Behavioural loyalty is repeat purchase; emotional loyalty is repeat preference even when an equal or better alternative is available. A customer can show up on every behavioural metric — frequency, recency, basket size — while harbouring zero preference for your brand over the next one. Strip away the discount, the points multiplier or the geographic convenience, and behavioural loyalty often dissolves instantly. Emotional loyalty doesn't, because it isn't contingent on the deal.

This distinction isn't new in principle — Frederick Reichheld and W. Earl Sasser Jr. made the economic case for it in their 1990 Harvard Business Review article "Zero Defections: Quality Comes to Services", which showed that modest improvements in customer retention produced outsized gains in profitability because loyal customers cost less to serve, refer others, and tolerate price increases. What that research didn't have, in 1990, was the behavioural-economics vocabulary to explain why retention and emotional attachment diverge from transactional repeat purchase. We have that vocabulary now, and it points to a hard truth: a programme can retain customers through friction and loss aversion while actively failing to earn their preference. Retention without preference is a liability waiting for a better offer to arrive.

Noon's loyalty approach in the Gulf is instructive here, precisely because it resists the instinct to lead with discounting. As we explored in our look at Noon's CX strategy, trust-building moves — reliable delivery promises, transparent returns, consistent service recovery — tend to produce preference that survives a competitor's price cut. A points balance rarely does.

What should you actually measure instead of points?

None of this means scrap the points ledger — it's useful operational data. It means stop mistaking it for the whole picture. A sound loyalty measurement model layers behavioural data with attachment data. The metrics that matter most sit in five categories:

• Share of wallet, not just share of visits. What percentage of the customer's total category spend comes to you, versus competitors? A customer who visits weekly but splits spend evenly across three brands is not loyal in any meaningful sense — they're diversified.
• Price elasticity of retention. Track what happens to repeat-purchase rate when you remove or reduce the incentive for a defined cohort. If retention collapses without the discount, you've been measuring mercenary behaviour all along.
• Voluntary advocacy, measured by actual referral behaviour — not a survey score, but tracked referral links used, reviews written unprompted, or branded content shared without an incentive attached. Advocacy that has to be paid for with a referral bonus is a transaction, not loyalty.
• Switching resistance. When a competitor runs an aggressive acquisition offer in your category, what share of your customer base stays put despite the better headline deal? This is the single cleanest proxy for emotional loyalty available, because it isolates preference from inertia.
• Customer lifetime value trajectory, not a snapshot. A single CLV number tells you what a customer is worth today. The trajectory — is CLV per cohort rising, flat, or declining as tenure increases — tells you whether the relationship is deepening or merely persisting.

Rakuten's ecosystem-based loyalty model offers a useful real-world illustration of measuring beyond the transaction. As detailed in our analysis of what Rakuten gets right, the company's cash-back points system is only one layer; the deeper retention driver is how embedded the ecosystem becomes across a member's daily purchasing decisions — a proxy for share of wallet that a simple points-redemption count would never surface.

How does the peak-end rule change what you should measure, and when?

Most loyalty programmes measure continuously — every transaction logged, every point tallied — on the assumption that more data points produce a more accurate loyalty picture. Behavioural science suggests the opposite: customers don't remember a relationship as an average of every interaction. Daniel Kahneman, Barbara Fredrickson, Charles Schreiber and Donald Redelmeier demonstrated this in their 1993 study "When More Pain Is Preferred to Less", published in Psychological Science: people judge an experience almost entirely by its most intense moment and its ending, largely disregarding duration and averaging the rest. This is the peak-end rule.

Applied to loyalty measurement, the peak-end rule means two things most dashboards ignore:

• The moments that determine whether a customer re-ups, renews or leaves are disproportionately the service-recovery moments — a flight disruption handled well, a billing error resolved generously, a complaint met with genuine empathy — not the average of every ordinary transaction. A loyalty model that only tracks steady-state satisfaction will miss the handful of moments actually deciding retention.
• How a relationship ends, or appears to end, carries outsized weight. A subscription cancellation flow, an account-closure experience, or even how a loyalty tier downgrade is communicated shapes the customer's retrospective verdict on the entire relationship — including whether they'll come back or recommend you to others.

This is why a loyalty measurement model needs a layer that specifically scores emotional peaks and endings, not just rolling averages. It's also why service recovery and escalation handling deserve to sit inside the loyalty conversation rather than being treated as a separate complaints function — see our thinking on escalation strategy for how that connects operationally.

How do you build a loyalty measurement model that goes beyond points?

Building this isn't a one-off survey project — it's a structural change to how loyalty gets tracked and reported. The sequence below is the one we use with clients moving off a points-only scorecard.

1. Separate behavioural data from attachment data in your reporting. Points issued, redeemed and tier status go in one bucket labelled "transaction activity." Share of wallet, referral behaviour, and switching resistance go in a separate bucket labelled "preference strength." Never let the two blend into a single composite "loyalty score" — it hides exactly the divergence you're trying to see.
2. Run a controlled incentive-withdrawal test. Select a representative cohort, suspend or reduce their points multiplier for a defined period, and measure what happens to repeat-purchase frequency relative to a control group. The delta is your mercenary-loyalty exposure, quantified in real numbers rather than assumption.
3. Instrument the moments, not just the month-end. Identify the three or four highest-stakes touchpoints in the customer journey — typically onboarding, first complaint, first renewal decision, and any account-closure or downgrade flow — and measure satisfaction and emotional response specifically at those moments, in addition to any ambient tracking.
4. Track voluntary advocacy as a distinct KPI, reported monthly. Unprompted reviews, organic social mentions, and referral-link usage without an active incentive campaign running are a cleaner proxy for genuine preference than any survey question, because they cost the customer effort with no points payoff attached.
5. Model CLV trajectory by cohort, not by individual. Group customers by acquisition channel and tenure band, then watch whether CLV rises or flattens as each cohort ages. A flattening trajectory, even with stable headline retention, is an early warning that loyalty is behavioural rather than emotional.
6. Report switching resistance after every major competitor promotion. Whenever a competitor runs a notable acquisition offer in your category, measure the defection rate among your base in the following 60 to 90 days. This single number, tracked over time, is the most honest loyalty metric most businesses never bother to capture.

Running this model requires marrying financial modelling with experience data — which is precisely where a lot of loyalty programmes stall, because the finance team owns the points liability and the CX team owns the satisfaction scores, and the two rarely sit in the same spreadsheet. Our CX ROI Calculator is a useful starting point for putting a number on what emotionally loyal cohorts are actually worth relative to mercenary ones, before you commit budget to rebuilding the measurement stack.

What role does reciprocity play in loyalty that points can't buy?

One of the more reliable, if underused, levers in loyalty design is reciprocity — the instinct, documented across decades of social-psychology research, to return a favour that wasn't transactional in the first place. A surprise upgrade, a handwritten note, a fee waived without the customer asking: these gestures work precisely because they sit outside the points economy. The customer can't "earn" them through volume, which is exactly why they generate goodwill a tier-based redemption never will. We've written at length about how small, non-transactional gestures build attachment in ways a points ledger structurally cannot replicate, in our piece on how small gestures build customer loyalty through reciprocity.

The design implication is straightforward: if every positive gesture in your programme is point-denominated, you've built a vending machine, not a relationship. The gestures that move the emotional-loyalty needle are, almost by definition, the ones that weren't priced.

How should loyalty programmes be redesigned around this measurement shift?

Measurement without redesign just produces a more honest version of the same disappointing number. Once a business can see the gap between behavioural and emotional loyalty, the programme itself usually needs structural change, not a cosmetic refresh of the points table. A few design moves follow directly from the metrics above:

• Weight tier status partly on advocacy and tenure, not purely on spend. A customer who has stayed five years and referred three friends carries more retention value than a high-spend newcomer, even if their points total is lower.
• Build recognition moments that can't be redeemed, only felt. Status that confers genuine human treatment — being remembered by name, having a prior issue proactively followed up on — creates endowment-style attachment that a points balance, however large, does not. Richard Thaler's foundational 1980 paper "Toward a Positive Theory of Consumer Choice", published in the Journal of Economic Behavior & Organization, is the origin of the endowment effect: people value what feels like theirs far more once they possess it. A relationship a customer feels they own is far stickier than a balance they merely hold.
• Protect the ending. Given the peak-end rule's weight on how experiences conclude, invest disproportionately in downgrade, pause and cancellation flows. A customer leaving gracefully, with dignity and an open door, is far more likely to return than one who feels punished on the way out.

None of this is a case against points. Points are an efficient mechanism for driving frequency and funding a rewards economy. The mistake is letting the points ledger stand in as a proxy for the customer relationship itself, when it was only ever built to measure the transaction layer. For a broader view of how loyalty strategy fits within a wider CX operating model, our customer loyalty consulting work looks specifically at separating retention mechanics from genuine preference-building, and our thinking on customer experience strategy covers how to sequence that redesign without disrupting an existing programme's commercial performance.

The number that should worry you more than redemption rate

If you want one number to chase above all others, it isn't points issued, points redeemed, or even headline retention. It's the defection rate of your base in the 90 days after a competitor's best acquisition offer lands. That number has nowhere to hide. It can't be flattered by a generous points multiplier or a sign-up bonus, because by definition the competitor's offer is already on the table. Everyone who stays despite it is loyal in the only sense that will ever matter commercially — and everyone who leaves was never yours to begin with, however full their points balance looked on the dashboard.