How to Measure Brand Loyalty Effectively

Loyalty reporting tends to drift toward whichever number is easiest to pull. For a lot of teams that means NPS, because the survey tool ships with it. Teams with no surveys fall back on repeat orders, since the order table is right there, and that has its own problem.

Neither one is wrong exactly. A repeat purchase records habit as readily as preference, and a survey score records intent with nothing bought behind it. Measuring brand loyalty means reading the two together, then converting the result into revenue through customer lifetime value.

NPS is the clearest example of leaning too hard on one half. Reichheld’s 2003 Harvard Business Review article reported Net Promoter Score as the best or second-best predictor of growth in 11 of 14 industries. MeasuringU’s review of that data found the scores were matched to growth that had already happened, which establishes correlation and not prediction.

Behavioral loyalty versus attitudinal loyalty

Purchase data shows how reliably customers choose a brand. Survey data shows how strongly they prefer it. The two disagree more often than teams expect. A customer can rebuy every month out of habit and feel nothing for the brand, or admire a brand and rarely buy because the category turns over once a decade.

Behavioral loyalty

The purchase record holds it. That means orders, renewals, returns, subscription continuity, and the share of category spend that goes to one brand.

Jacoby and Chestnut (1978) defined brand loyalty as a biased behavioral response, expressed over time, toward one or more brands in a set, and driven by psychological processes. The “biased” part is the catch, since a repeat purchase with no preference behind it is not loyalty by their definition.

Transaction data records the act and never the reason for it.

Attitudinal loyalty

This covers what customers say they feel and what they say they will do. Surveys capture it through recommend likelihood, repurchase intent, brand preference and brand commitment. Scores arrive before behavior does, so they work as an early signal and a weak proof.

How weak got tested directly. A two-year panel of more than 8,000 US customers across retail banking, mass-merchant retail and internet service providers let Keiningham, Cooil, Aksoy, Andreassen and Weiner (Managing Service Quality, 2007) compare stated intent with later behavior. Recommend intention alone did not predict loyalty behavior well enough to stand as a single measure, while a set of several indicators predicted retention and recommendations better.

Composite and stage models

Composite approaches score both dimensions together, so a customer who buys often and rates the brand highly outranks one who does only one of the two.

Aaker’s loyalty pyramid (1991) sorts customers into five levels, from least to most loyal:

1. Switchers, who buy on price or availability
2. Habitual buyers, who are satisfied or at least not dissatisfied
3. Satisfied buyers with switching costs
4. Brand likers, who have an emotional attachment to the brand
5. Committed buyers

Levels two and three look identical in a sales report. Only a survey or a price test separates habit and switching costs from preference.

How to calculate behavioral loyalty metrics

Repeat purchase rate, retention and churn, purchase frequency and share of wallet do most of the work on the behavioral side. All but share of wallet run on first-party transaction data. Share of wallet needs competitor spend, which only a survey or a purchase panel can supply.

Repeat purchase rate

Divide the customers with two or more orders in the period by all customers in the period, then multiply by 100.

Take a store with 1,000 customers and 1,400 orders in 12 months, where 240 of those customers ordered at least twice (hypothetical numbers). The repeat purchase rate is 24%.

Some teams count orders instead of customers. On the same data, repeat orders (1,400 minus 1,000 first orders, so 400) are 28.6% of all orders, a different number answering a different question.

Say which method a report uses, and keep the window close to the category’s normal repurchase cycle. A 30-day window undercounts loyal buyers of anything purchased twice a year.

Customer retention rate and churn rate

For retention, take the customers you have at the end of a period, remove anyone acquired during that period, and divide by the customers you had at the start. Then multiply by 100.

Written out, that is ((customers at end of period minus new customers acquired) divided by customers at start of period) x 100.

With 500 customers at the start and 540 at the end, 90 of them new, the sum is (540 – 90) / 500 = 90% retention. Churn rate is the remainder, 10%.

Define “active” before counting. In a non-subscription business, a customer with no purchase in the window has not formally left, so the rule for a lapsed customer (for example, no order in 12 months) moves the result more than the arithmetic does.

Purchase frequency

Divide total orders by unique customers for the same period. The example store above scores 1.4 (1,400 orders across 1,000 customers).

The average hides skew. A few heavy buyers pull it up while most customers order once, so report the median next to the mean.

Purchase frequency also feeds customer lifetime value, where it multiplies against order value.

Share of wallet

Brand spend divided by total category spend, times 100, gives the percentage of a customer’s category spend that goes to your brand.

Transaction data can’t see it. The figure comes from survey questions about how spend splits across brands, or from a purchase panel.

Recommend scores can’t stand in for it either. A customer can rate two competing brands 9 out of 10 and still split spend evenly between them.

How NPS, CSAT and CES measure attitudinal loyalty

Net Promoter Score (NPS) tracks advocacy and customer effort score (CES) tracks friction, while customer satisfaction score (CSAT) grades one specific experience. Each rests on a single question, which makes them cheap to run and easy to misread.

Net Promoter Score

Frederick Reichheld of Bain & Company introduced NPS in a December 2003 Harvard Business Review article, “The One Number You Need to Grow.”

Customers who answer 9 or 10 are promoters, 7 or 8 are passives, and 0 to 6 are detractors. NPS is the percentage of promoters minus the percentage of detractors, so it runs from -100 to +100. Passives count toward the total but not the subtraction, which means 55% promoters and 20% detractors give a score of 35.

Charles Schwab is the showcase case. Bain credits a Net Promoter system as a key part of the firm’s turnaround after 2004, a period in which its stock tripled over four years.

The predictive claim has drawn empirical challenge. A 2007 Journal of Marketing paper by Keiningham, Cooil, Andreassen and Aksoy tested Reichheld’s argument that recommend likelihood is the one survey item tied to growth, and disputed it.

Most Net Promoter surveys add an open follow-up asking why, since the number alone doesn’t say what to fix.

CSAT and CES

CSAT looks like a loyalty measure and isn’t one, and the best evidence comes from a study of more than 75,000 customer service interactions. Dixon, Freeman and Toman (Harvard Business Review, 2010) found that 20% of “satisfied” customers intended to leave, while 28% of “dissatisfied” customers intended to stay.

That gap is why CSAT cannot stand in for loyalty. It grades one experience, and a good grade doesn’t predict whether the customer comes back.

It still earns a place as quality control at specific journey moments. CES suits problem-solving contacts such as support tickets, where what matters is how hard the customer had to work to get something done.

The same authors concluded that reducing effort builds loyalty, while delighting customers does not.

Survey question wording and timing

A relationship survey goes out on a schedule to a sample of the whole customer base and asks about the brand overall. Transactional surveys fire after a purchase or support contact and cover only that event. Mixing the two corrupts the trend line, so keep them apart.

Sample across the whole base, not only recent buyers, or the score reflects whoever bought last week.

Beyond the headline scores, a few more questions fill in the attitudinal picture:

• Repurchase intent, asked as “How likely are you to buy from us again in the next six months?”
• Brand preference, asked as “Which brand would you choose first in this category?”
• Switching tendency, which is “If we were unavailable, what would you do?”
• Advocacy, asked as “Have you recommended us to anyone in the last year?”

The last question asks about something the customer did, not something they might do. That makes it a useful check on stated intent.

Customer lifetime value and loyalty program ROI

CLV is where loyalty becomes a revenue figure. Simple CLV multiplies average order value, purchase frequency and customer lifespan. Loyalty program ROI is the separate question of whether a program earned more than it cost.

With a $60 average order, 1.4 orders per year and a 3-year lifespan (hypothetical numbers), CLV comes to $252. At a 40% gross margin, margin-adjusted CLV is $100.80.

Lifespan is the weakest input. A common shortcut sets it at 1 divided by annual churn, so 10% churn implies 10 years, and that extrapolation breaks when the brand has only a year or two of data.

For program ROI, subtract program cost from incremental revenue, divide by program cost and multiply by 100. The word that matters is incremental. Revenue members would have generated anyway does not count. Retailers weighing that question can study Open Loyalty’s breakdown of retail loyalty mechanics to see which ones move repeat purchases in their category.

The quick way to test this is to compare members with non-members. It’s fast to run and easy to explain, and it overstates the effect, because customers who expect to benefit most enroll first.

A randomized holdout works better. A small random group is excluded from the program or from one campaign, and you compare purchase frequency, basket size and repeat purchase rate over at least one full purchase cycle.

The holdout is slower and sometimes awkward to sell internally. It is also the comparison a finance team is least likely to pick apart.

How do RFM analysis and cohort tables segment loyal customers?

Scoring every customer on recency, frequency and monetary value (RFM) shows who the loyal ones are today. A cohort retention table, which follows groups of customers by acquisition period, shows when everyone else drifted away.

Recency is days since the last purchase, and fewer days score higher. Frequency counts orders in the window, and monetary value is total spend in that same window.

Each dimension is commonly ranked from 1 to 5 by quintile, and the scores combine into segments with labels such as champions, loyal customers, at risk and hibernating.

Quintiles rank customers against each other, so a 5 in a low-frequency category is not absolute loyalty. Watch for frequent buyers whose recency score is falling, because they are the lapsing loyal customers.

In a cohort retention table, rows are acquisition months, columns are periods since the first purchase, and each cell shows the share of that cohort still buying. A curve that flattens means a loyal core has formed. A curve that keeps falling means customers try the brand and leave.

Google Analytics 4 builds this view in its cohort exploration report. Per Google’s Analytics Help documentation (checked September 2026), you set an inclusion criterion that puts a user into a cohort, such as acquisition date, and a return criterion that the user must meet again later, at daily, weekly or monthly granularity.

When you break cohorts down by a demographic dimension, Google applies data thresholds: cohorts too small to protect user anonymity are left out of the exploration, so small segments can drop out of the table.

Segment-level reporting matters because a blended average can hide a loyal core and a large group of one-time buyers.

Which brand loyalty measurement approach fits your business?

Pair one behavioral metric with one attitudinal metric and one value metric, and set the tracking window to the category’s repurchase cycle.

The lead metric changes with the business model. This table shows the default pairing for four common cases.

Low-frequency categories like furniture, cars and mattresses should lead with intent and preference questions, because the next purchase is years away and behavior gives no signal in between. Grocery, coffee and meal delivery run the other way. Behavior leads there, and surveys explain the movement.

A few hundred customers make cohort tables too thin to read. At that size, review accounts one by one.

Window choice is where most setups go wrong. Amplitude’s retention guide makes the point with a contrast: daily usage matters for a social game, but a food delivery app may only expect an order once or twice a month.

The obvious default is NPS, because it’s one question and every vendor supports it. For most brands a retention or repeat purchase metric should lead and NPS should explain it, for reasons the limits section below covers.

Loyalty data sources, tools and tracking setup

Transaction data, CRM records, loyalty program data, surveys and social listening all feed loyalty measurement. Each covers part of the picture and misses another part.

Data sources and what each misses

• Transaction data comes from the order system or point of sale (orders, basket size, dates). It misses guest purchases and all competitor spend.
• CRM records hold account and contact history in Salesforce or HubSpot, and they are only as clean as the deduplication behind them.
• Loyalty program data covers points, tiers and redemptions from platforms such as Yotpo and Smile.io, for members only.
• Surveys (NPS, CSAT, CES and intent questions) cover responders only.
• Social listening picks up public mentions and sentiment, so quiet loyal customers never show up.

No single source shows both what customers do and why. Joining them on one customer ID is the real work.

Tools by job

Mixpanel’s retention report counts a customer once per bucket but can place the same customer in several, so a weekly buyer appears in every weekly bucket. Amplitude offers three definitions (return on, return on or after, and custom brackets) that give different numbers from identical data.

Yotpo’s documentation describes dashboards that compare active with inactive loyalty members. Smile.io’s help center defines redemption rate as points redeemed divided by points earned over the last 30 days, a fixed window worth knowing before you compare it with anything else.

Setup steps

1. Define the loyalty event (any purchase, a specific product, a renewal) and the rule for a lapsed customer.
2. Unify customer IDs across the order system, CRM and loyalty platform so one person is one record.
3. Set the tracking window from the category’s repurchase cycle.
4. Baseline repeat purchase rate, retention rate and purchase frequency before changing anything.
5. Schedule relationship and transactional surveys, sample across the whole base, and link each response to a customer ID.
6. Report every metric by segment and cohort, not only as one blended figure.

Step two is where setups tend to stall. Guest checkouts and separate loyalty IDs split one customer into several records.

What is a good loyalty benchmark by industry?

A good score depends on who built the benchmark. Qualtrics XM Institute’s 2024 consumer study put its best industry average at about 34, while Retently’s 2025 data put insurance at 80.

• Qualtrics XM Institute (2024) had 10,000 US consumers rate 354 companies across 22 industries. Grocery scored highest at about 34, car rental lowest at 15.8.
• SurveyMonkey (2025) reports a global average NPS of +32 and a median of +44, across more than 150,000 organizations.
• Retently’s 2025 benchmark has insurance averaging 80 and ecommerce and retail 59. Its 2026 edition puts insurance at 46.
• The ACSI 2025 retail study scored the retail sector 78.3 on a 0 to 100 satisfaction scale.

Put the three NPS sources side by side and one score reads three ways. An NPS of 45 beats every industry average in the Qualtrics XM Institute data, sits near the median in SurveyMonkey’s, and trails Retently’s ecommerce and retail figure by 14 points.

The likely cause is sample design. Retently’s figures come from its own customers, companies that already run feedback programs, while Qualtrics XM Institute asks a consumer panel to rate companies they recently dealt with.

Retently’s insurance average fell 34 points in one year. That looks more like a change in who was surveyed than a change in how insurance customers feel.

ACSI is a national cross-industry index that measures satisfaction, not loyalty, on a 0 to 100 scale. It is not interchangeable with NPS, which runs from -100 to +100.

The Brand Keys Customer Loyalty Engagement Index surveys consumers each year on over a thousand brands and ranks them within categories on drivers it says predict engagement over the next 12 to 18 months. The 2025 edition awarded Levi Strauss gold in apparel and Macy’s gold in department stores, which tells you a rank, not a score to compare against your own.

Retention and repeat purchase benchmarks are the weakest category. They circulate mostly through vendor blogs without a stated method, so treat them as unverified.

Benchmark against your own trend first. A method held constant across survey waves is worth more than any cross-industry average.

When do loyalty metrics mislead?

Inertia loyalty, survey response bias, NPS predictive gaps and short tracking windows cause most false readings.

Inertia loyalty

Dick and Basu (Journal of the Academy of Marketing Science, 1994) sorted customers into four groups by relative attitude and repeat behavior. Their “spurious loyalty” group, with high repeat purchasing and a weak attitude, is inertia.

The typology is among the most cited in loyalty research, yet Garland and Gendall (Australasian Marketing Journal, 2004) note that Dick and Basu never operationalized it or tested its predictive power. An earlier supermarket study by East, Sinclair and Gendall (2000) found little support for it, so use it as a diagnostic lens and not as a classifier you can trust.

Survey response bias

Scores describe responders, and responders are a self-selected slice of the customer base. Rocco (2019, Springer) ties the growing concern to declining response rates and treats the distortion as a self-selection effect.

The damage comes from how responders differ from non-responders, not from the response rate alone. A 40% response rate that misses every lapsed customer is worse than a 10% rate that includes them.

NPS predictive gaps

Zaki, Kandeil, Neely and McColl-Kennedy (Cambridge Service Alliance working paper, 2016) combined behavioral and attitudinal data in a B2B setting and confirmed that NPS does not necessarily correspond to actual behavior. Their model also reads customers’ verbatim comments to explain why accounts churn.

Sources disagree here. Bain, which co-developed the Net Promoter system, states that on average an industry’s loyalty leader grows more than twice as fast as its competitors. The critiques that Zaki and colleagues summarize in their 2017 Cambridge Service Alliance webinar report no empirical evidence that attitudinal loyalty metrics correlate significantly with relative revenue change within an industry.

The sources do not settle it, so test it on your own data. Join each customer’s score to what that customer did next.

Short tracking windows

Amplitude’s documentation warns that retention curves on ongoing data can trend upward, because each point counts only users who have had enough time to reach it. The same app can also show noticeably different retention depending on whether the report uses an exact-day or an on-or-after definition.

A window shorter than the purchase cycle undercounts loyal customers, and an unfinished period flatters the curve.

Brand Loyalty Measurement FAQ

How is brand loyalty different from customer satisfaction?

Satisfaction grades an experience. Loyalty is a repeated choice of one brand over its rivals.

A satisfied customer can still leave, so treat CSAT as one input and track repurchase and retention as the outcome.

Can you measure brand loyalty without a loyalty program?

Yes. Order history alone yields repeat purchase rate, retention rate and cohort tables.

The catch is identity. Each order needs an account or email attached, because guest checkouts cannot be tied back to a single customer record.

How often should you measure brand loyalty?

Review behavioral metrics monthly and run relationship surveys quarterly or twice a year.

Judge any trend over at least one full purchase cycle. A single month of data in a slow-repurchase category proves nothing.

Does social media engagement count as a loyalty metric?

Only as a supporting signal. Likes, comments and shares measure the vocal part of an audience, and many loyal customers never post.

Pair engagement with social listening sentiment, and never let it replace repurchase data.

What should you measure when a brand has only a few hundred customers?

Skip percentage-based dashboards. With so few customers, one lost account swings every rate.

Track renewals and reorders account by account, and ask customers directly why they stayed or left.

Which Signal to Trust When Loyalty Metrics Disagree

In brand loyalty measurement, behavior outranks stated intent when the two conflict. A falling retention rate beside a rising NPS means customers are leaving while still saying they would recommend the brand.

Start with the cohort retention table, then look at the survey responder mix, then read the verbatim follow-up comments. A lapsed customer costs revenue and a skewed survey only misleads a report, which is why the order runs that way.

Survey swings deserve the second check. Retently’s insurance average fell 34 points in a year, equal to Qualtrics XM Institute’s best industry average of about 34, which points to a change in who answered.

The cost of this order is acting later, since a retention drop confirms only after a full purchase cycle closes.

This order holds as of September 2026, and a holdout test showing survey scores predicting repurchase would change it. Widening the view to overall brand performance is the next step.

• Author
• Recent Posts

Bogdan Sandu

Bogdan Sandu is a seasoned designer who has been designing websites since 2008.

Renowned for his expertise in logo design and visual branding, Bogdan has developed a multitude of logos for various clients.

His skills extend to creating posters, vector illustrations, business cards, and brochures. Additionally, Bogdan's UI kits were featured on marketplaces like Visual Hierarchy and UI8.

He also wrote in the past years on sites like Design Your Way, WebDesignerDepot, WPDean, Designmodo, Speckyboy, Slider Revolution, and more.

Latest posts by Bogdan Sandu (see all)

• How to Unlock Canva Page and Locked Elements - 4 November 2026
• How to Design a Google Docs Template That Looks Professionally Designed: Fonts, Grid and Color - 4 November 2026
• Black Metal Font Generator - 2 November 2026