Customer loyalty has two halves, and they often disagree. One half is behavior: what customers buy and how often. The other half is attitude: how they rate the brand against the other options they use. Read one half alone and the numbers mislead.
The gap is wide. In a three-year study of more than 75,000 customers, 20% of satisfied customers said they planned to leave. Of the dissatisfied ones, 28% planned to stay. Oliver (1999) reached the same view from theory. Loyal customers are usually satisfied, but satisfaction does not always turn into loyalty.
What are the three types of customer loyalty metrics?
Loyalty metrics fall into three groups. Behavioral metrics come from sales data. Attitudinal metrics come from surveys. Program metrics come from the loyalty platform and from controlled tests. The table defines the core metric in each group and the blind spot each one carries.
| Type | Metric | How to calculate it | What it misses |
|---|---|---|---|
| Behavioral | Repeat purchase rate | Customers with two or more purchases in the window divided by customers with at least one | Habit and deal seeking look the same as preference |
| Behavioral | Purchase frequency | Purchases divided by active customers in the window | Promotions pull spend forward and inflate it |
| Behavioral | Share of wallet | Your sales to a customer divided by that customer's total spend in the category | The category total needs a survey or panel |
| Behavioral | Cross-use | Customers using two or more of your services divided by all customers | Breadth can hide thin spend in each service |
| Attitudinal | Net Promoter Score | Percentage of promoters (9 or 10) minus percentage of detractors (0 to 6) | Stated intent, self-reported and easy to game |
| Attitudinal | Satisfaction | Satisfied responses divided by all responses | Satisfied customers still leave |
| Attitudinal | Customer Effort Score | Rating of the effort to resolve an issue, 1 to 5 in the original HBR scale | Covers service contacts only |
| Attitudinal | Emotional connection | Panel survey on what the brand means to the customer, linked to spend | Moves slowly and needs a standing panel |
| Program | Active rate | Members with a set action in the window divided by all members | Shifts whenever the window or the action changes |
| Program | Redemption rate | Points redeemed divided by points issued | High redemption can mean the program pays for spend that would happen anyway |
| Program | Incremental lift | Spend per customer in the test group minus spend per customer in a random holdout | Needs a clean holdout and enough time |
How do you measure behavioral loyalty?
Start with identification. You can only measure behavior on sales tied to a known customer. Kroger says its rewards program puts a loyalty card on over 95% of customer purchases, per its annual report. A program that knows the buyer on half its sales sees half its customers.
Next, fix the time windows. Repeat rate and purchase frequency need a set period that matches the buying cycle, such as 90 days for groceries or a year for travel. Recency, the days since the last purchase, flags a lapse early. Keep every definition the same from one report to the next.
Share of wallet is the hardest behavioral metric to get. The HBR authors who studied it call it what matters most. It is the percentage of a customer's spending in a category that one brand captures. Your own data gives the top of that ratio, and the category total has to come from a survey or a panel. Hilton's 2025 annual report shows the payoff: the share of travel spend it captures rises as Honors members move up the tiers.
Breadth matters for any business with more than one line. Rakuten reports the share of its users on two or more of its services. Over the 12 months to June 2026, it was 77.0%, per its quarterly results. For a membership, renewal rate does the same job. Costco reports 92.2% in the U.S. and Canada and spells out the method: a trailing count of renewals from seven to eighteen months before the report date.
Read every behavioral number against brand size. Under the pattern researchers call double jeopardy, a small brand has far fewer buyers than a large one, and those buyers buy it less often. A smaller brand's lower repeat rate can be normal for its size. Compare it with brands that have a similar number of buyers.
How do you measure attitudinal loyalty?
Net Promoter Score. Fred Reichheld set it out in Harvard Business Review in 2003. Customers rate how likely they are to recommend the company to a friend or colleague, from 0 to 10. Per Bain, the score is the percentage of promoters, who answer 9 or 10, minus the percentage of detractors, who answer 0 to 6.
NPS is quick to run, and its strongest claim has not held up in every test. A 2007 Journal of Marketing study drew on 21 firms and more than 15,500 interviews. In industries Reichheld had cited, it failed to replicate the claim that Net Promoter was clearly better than other measures.
Satisfaction and rank. Satisfaction shows whether a customer is content. It says little about where the next dollar goes. In a two-year study of more than 17,000 consumers, changes in satisfaction or intent to recommend tracked changes in share of wallet at a correlation of only 0.1. Changes in a score built on the rank a customer gives the brand tracked it at 0.8. So add one question to every survey: how does this brand rank against the others you use?
Customer Effort Score. The HBR study of 75,000 customers introduced it. Customers rate the effort it took to get an issue solved, on a scale of 1 to 5. The authors found it predicted loyalty better than satisfaction or Net Promoter Score did. Ask it right after service contacts.
Emotional connection. Emotional loyalty is the bond that holds when a rival matches the price. HBR research traces a path in four steps. Customers start unconnected, become highly satisfied, then see the brand as different, then feel fully connected. Fully connected customers were 52% more valuable on average than highly satisfied ones. Measure it with a short panel survey on what the brand means to each member, linked to their record.
How do you measure what the loyalty program adds?
Program metrics show whether members use the program. The active rate counts members with a set action in a set window, such as a purchase in 90 days. The redemption rate divides points redeemed by points issued. Both describe use. Neither one shows cause.
Members choose to join, and customers who already buy often have the most reason to sign up. A member premium therefore mixes the program's effect with who joined. A holdout pulls the two apart. Keep a random group of eligible customers out of an offer, then compare spend and margin per customer. Our breakdown of loyalty program KPIs runs a holdout step by step. The loyalty program ROI guide prices the result.
How do you combine behavioral and attitudinal loyalty?
Put both readings on the same member record. Send the survey to a sample of known members, then join each answer to that member's purchases. Score behavior high or low on share of wallet or frequency. Score attitude high or low on brand rank or NPS. Four groups follow, and each one calls for a different budget.
- Committed: buy often and rank you high. Thank them with recognition and good service. Hold discounts back, since they buy anyway.
- Habitual: buy often but rank you no higher than rivals. A rival's matching offer puts this spend at risk, so watch this group first.
- Untapped: rank you high but buy elsewhere. Find the barrier, whether it is price, range, location or channel.
- Uncommitted: low on both counts. Spend here only where a holdout shows lift.
For illustration, a retailer surveys 10,000 known members and joins each answer to a year of spend. Of those, 3,000 land in the habitual group, spending $800 a year each. That is $2.4 million a year, in the sample alone, that rests on habit and price. Another 1,500 rank the brand first but spend $200 a year. Closing their $600 gap to the habitual group is worth $900,000 a year. The first group needs defending. The second needs a barrier removed.
What should a loyalty measurement plan include?
A good plan ties every metric to a decision. For each question, name the metric, the data source, how often it runs and who owns it. Then write down what the team will do when the number moves. The plan below fits an enterprise program. Set each cadence to your own buying cycle.
| Question | Metric | Source | Cadence | Owner | Decision it drives |
|---|---|---|---|---|---|
| Are customers coming back? | Repeat rate, recency, active rate | Identified sales | Weekly | CRM analytics | Trigger lapse offers |
| Are we winning the category? | Share of wallet, brand rank | Survey joined to sales | Quarterly | Customer insights | Move budget between segments |
| Do customers use more than one line? | Cross-use | Sales across business lines | Quarterly | Loyalty strategy | Pick the next partner or product to add |
| Would they recommend us? | NPS with reason codes | Relationship survey | Quarterly | Customer experience | Fix the top detractor themes |
| Was service easy? | Customer Effort Score | Post-contact survey | After each contact | Service operations | Remove repeat contacts and channel switches |
| How attached are they? | Emotional connection | Member panel | Twice a year | Brand and insights | Fund recognition and experiences |
| Is the program used? | Active rate, redemption rate | Loyalty platform | Monthly | Loyalty team | Refresh rewards and earn rules |
| What did the program cause? | Incremental lift and ROI | Random holdout | Every major offer, plus a long-run control | Loyalty finance | Fund, change or stop the offer |
What are the common mistakes in measuring customer loyalty?
Treating satisfaction as loyalty. In the HBR effort study, one in five satisfied customers planned to leave. Being satisfied is the floor.
Paying bonuses on NPS. Reichheld and two Bain colleagues wrote in 2021 that inexperienced users abused NPS by tying it to bonuses for frontline staff. Staff then cared more about the score than about serving customers better. Use the score to learn. Pay on outcomes such as retention or share of wallet.
Reading member share of sales as growth. A rising member share of sales shows reach. Ulta draws about 95% of its sales from members, per its annual report. That figure describes coverage. Only a holdout shows what the program caused.
Moving the definitions. Stretch the active window from 90 days to 12 months and the active rate jumps while behavior stays the same. Costco shows the fix. It publishes how its renewal rate is calculated, and it names the mix effects that moved it.
Comparing brands of different size. Under double jeopardy, a smaller brand's buyers buy it less often. Benchmark repeat rate against brands with a similar number of buyers.
Keeping surveys and sales apart. When attitude scores sit in one system and spend sits in another, the two never meet. Join them at the member level. Until then, the habitual and untapped groups stay hidden.
How GRAVTY handles loyalty measurement
GRAVTY keeps what members do and what the program does on one record. Member 360 puts each member's behaviors and sales in one view, and it shows the whole household. A patented method, Multi-Dimensional Behavior Tracking, turns what members do into first-party data. Compass, the GRAVTY agentic AI suite, reads program KPIs next to sentiment. It flags odd spikes and dips and compares results to benchmarks. It simulates an offer's outcome before launch. Each of its suggestions can be checked and explained.
Frequently asked questions
What is the best way to measure customer loyalty?
Pair one behavioral metric with one attitudinal metric on the same customer record. Share of wallet with brand rank is a strong pair, because changes in rank track changes in share of wallet closely. Add a holdout test whenever you need to know what the loyalty program caused.
How do you calculate a customer loyalty rate?
Calculate it as the repeat purchase rate: customers who bought two or more times in a window, divided by customers who bought at least once. Fix the window to your buying cycle, such as 90 days for groceries or a year for travel.
Is Net Promoter Score a good measure of customer loyalty?
NPS is a good measure of stated advocacy and a weak guide to share of wallet on its own. In one two-year study, changes in intent to recommend tracked changes in share of wallet at a correlation of only 0.1. Pair it with purchase data.
How often should you measure customer loyalty?
Measure behavior weekly or monthly, attitude each quarter and emotional connection twice a year. Run a holdout with every major offer. Keep a small long-run control group for changes to tiers or earn rates.
What data do you need to measure customer loyalty?
You need sales tied to known customers, plus a survey you can join to member IDs. A random holdout adds the program's effect. Know the buyer first. Kroger ties over 95% of customer purchases to a loyalty card, so its sales data covers nearly every buyer.
Can you measure emotional loyalty?
Yes. Survey a panel of members on what the brand means to them and how it compares with rivals, then link each answer to their purchase record. In HBR research, fully connected customers were 52% more valuable on average than highly satisfied ones.