Most brand loyalty statistics in circulation trace back to a 1990 study of banks and insurance brokers, and the widely quoted 25 to 95 percent figure is a later restatement. The defensible 2026 numbers are lower, and AI agents are reshaping them now.
The single most quoted statistic in retention marketing was measured in a bank branch system the year Tim Berners-Lee invented the web. It has been restated, rounded up, and reattributed so many times that the version you know is not the version that was published.
74% of the customers who buy from you once will never buy again. That figure comes from Bluecore’s analysis of more than 100 retailers across apparel, beauty, footwear, home goods, and jewelry, and it is probably the most useful brand loyalty statistic in circulation right now. It is also not the one you have seen quoted.
The one you have seen quoted is that a 5 percent increase in retention lifts profits by 25 to 95 percent. It appears in vendor decks, agency pitches, LinkedIn posts, and roughly forty of the fifty pages currently ranking for brand loyalty statistics. I have quoted it myself. It is close to true, it is directionally useful, and the specific version everyone repeats is not what the underlying research said.
That matters more in 2026 than it did in 2023, for a reason that has nothing to do with academic tidiness. When a merchant searched Google for a statistic, they landed on a page and made their own judgment about whether it looked credible. When they ask ChatGPT, Claude, or Perplexity, the model returns a number stripped of its context, and the sourcing quality of the page it pulled from is invisible. Bad provenance now propagates silently and at scale. So this piece does something the rest of this category does not: it names the primary source, the sample, and the year for every number, and tells you plainly which ones to stop using.
The “25 to 95 percent” figure comes from a 2014 Harvard Business Review article, not from original research, and the study underneath it reported a range of 25 to 85 percent. The 2014 piece is Amy Gallo’s summary of customer churn economics, which credits Frederick Reichheld of Bain and Company. Gallo’s article is where the modern phrasing enters circulation, and it is also the origin of the companion claim that acquiring a customer costs five to 25 times more than retaining one, which she introduces with the qualifier “depending on which study you believe.”
Trace it one step further back and you reach Reichheld and Sasser’s 1990 paper on customer defection rates. The actual findings: cutting the defection rate by 5 percent generated 85 percent more profit in one bank’s branch system, about 50 percent in an insurance brokerage, and 30 percent in an auto service chain. Three service businesses. Named individually. In 1990.
None of that makes the finding wrong. The direction has held up across thirty-five years and the mechanism is sound, which is why it survived. But three things are worth holding onto. The ceiling was 85, not 95. The businesses studied were financial services and auto repair, not ecommerce, which did not meaningfully exist yet. And the range describes what happened at specific firms, not a law that applies to your store.
I want to be direct about something here. This site has published the 25 to 95 percent version, attributed to Bain, in more than one piece. That was the industry consensus and I repeated it without checking the chain. Checking it changed my view of how much of this category is built on restatement rather than research, and it is the reason this page now carries confidence ratings instead of a tidy list of percentages.
Eight statistics survive a provenance check, and one commonly repeated claim does not survive at all. The table below gives you the primary source, the year, and an honest confidence rating for each. Confidence reflects sample quality and independence, not whether the number is flattering.
That last row deserves a note. The 80/20 claim is the Pareto principle wearing a lab coat. It gets attributed to Bain, to Gartner, and to unnamed “studies,” depending on the page, and I could not find an underlying dataset for any version of it. Your own store either follows that distribution or it does not, and you can check in 10 minutes in Shopify Analytics. Use your number, not the folklore.
Across more than 100 retailers, the average repeat purchase rate is 16.5 percent, which is roughly ten points below the figure most Shopify content quotes. The Bluecore Customer Growth Benchmarks data puts health and beauty highest at 21.5 percent, followed by sporting goods and outdoor at 21.2 percent, and apparel at 20.2 percent.
The competing figure you will see everywhere is 28.2 percent, usually attributed to Shopify. Both can be right, because they measure different things over different windows and different store populations. Bluecore’s set skews toward larger omnichannel retailers over a full calendar year. That definitional gap is exactly why a category average is close to useless for your decision making, and why comparing yourself to brands at your price point and purchase cycle matters far more than beating an industry number.
The figure I would actually put in front of a board is the spend gap. Across every vertical, repeat buyers placed 57.6 percent more orders and spent 69.2 percent more than new buyers. That is transaction data rather than survey data, it is consistent across categories, and it makes the argument for retention investment without needing a 1990 banking study to prop it up. If you are building the system behind that number, the retention framework that moves repeat purchase rate covers the flow architecture in detail.
Loyalty and CRM now absorb 51.5 percent of the average marketing budget, and more than 90 percent of companies run a program, which means having one is no longer a competitive advantage. The Antavo Global Customer Loyalty Report 2026, based on 3,000 marketers and 10,000 consumers, also reports that 92.7 percent of program owners saw a positive return, averaging 5.3 times.
Read that ROI figure carefully. The respondents are the people who selected, bought, and are now accountable for the programs, surveyed by a loyalty platform vendor. That does not make it false, and the sample is large. It does mean the number carries a structural optimism that a neutral audit would not. I have marked it moderate confidence for that reason, and I would use it as evidence that programs generally work rather than as a forecast of what yours will return.
The more useful finding sits on the consumer side. LoyaltyLion’s 2026 consumer research found shoppers now belong to about three programs on average, up from 2.8 a year earlier, and that over half would stop participating if there were not enough ways to earn points beyond purchasing. Separately, Attentive’s survey of 600 US shoppers conducted in January 2026 found irrelevant recommendations and message overload actively push customers away.
Put those together and the pattern is one I saw constantly during my years at Shopify working with DTC brands: merchants at $500K to $2M install a loyalty app expecting it to fix retention, when the actual problem is that nothing happens between the first order and silence. A program layered on top of a broken post purchase experience adds a discount liability without adding a relationship. If your repeat rate is under 20 percent and you have no post purchase sequence, build the sequence first. When you are ready, the Shopify referral and loyalty apps worth evaluating in 2026 breaks down the options by stage, and if you would rather not build it in house, there are retention marketing agencies working with Shopify brands at the $5M and above tier.
Fifty-seven percent of consumers say they would let an AI shopping agent switch brands on their behalf if it found better value. That figure comes from Checkout.com’s June 2026 research on agentic commerce demand and merchant readiness, and it is the single number in this piece most likely to matter in eighteen months.
The supporting evidence is consistent. NIQ found that 42 percent of consumers used at least one AI tool to shop in the past month. McKinsey’s State of the Consumer 2026, surveying 4,863 consumers across five markets, found 28 percent of Gen Z already using generative AI for shopping and 60 percent regularly reading the AI overview at the top of search. Bain’s analysis of autonomous shopping frames the consequence in one line: AI shifts loyalty from brands to outcomes.
Here is the part that connects loyalty to visibility. McKinsey found that only 1 percent of the sources large language models cite when answering questions about a brand come from that brand’s own website. Even among the ten most cited sites, which account for roughly 22 percent of all citations, brand owned properties make up between 3 and 10 percent. Your product pages, your about page, your carefully written brand story: almost none of it is what the model reads when a shopper asks which brand to buy.
Apply the eighteen month test. Points and tiers still work, and I am not telling anyone to dismantle a program that is returning. But a loyalty program that only exists inside your checkout is invisible to an agent comparing options across a category. The brands that hold position are the ones making their offer, inventory, and reasons to choose them machine readable, which is what Shopify MCP exposes about your catalog to AI agents is for. Structured data is becoming a retention channel, which is a genuinely strange sentence to write and I think it is correct.
Name the primary source, the sample, and the year, or do not use the number at all. That test takes under two minutes and it disqualifies a startling share of what circulates in this category. If a page cites a statistic without a link, treat the statistic as unverified until you find the original.
Four failure modes account for almost everything that goes wrong. Range creep, where 25 to 85 becomes 25 to 95 through repetition. Attribution drift, where a number credited to a 2014 HBR summary gets recredited to Bain, then to Gartner, then to nobody. Context loss, where a finding from three service firms in 1990 gets presented as an ecommerce benchmark. And vendor framing, where a survey of people who bought a product reports how well that product works.
For your own store, the honest move is to stop benchmarking against category averages entirely. Pull your repeat purchase rate, your time to second purchase, and your revenue split between new and returning customers. Those three numbers tell you more than every statistic on this page combined, because they describe your business rather than an aggregate of businesses that are not yours. The connection between email, loyalty programs, and community and your retention numbers is where most of the movement comes from once you know your baseline.
I ran an online contact lens business before this, in a category where replenishment is the entire model. What I learned there has stuck with me through several hundred merchant conversations since: what looked like brand loyalty was mostly being the default reorder, and the moment a competitor made reordering marginally easier, the loyalty evaporated. That is worth remembering as agents get better at comparison. Convenience has always been doing more work than affinity, and we are about to find out exactly how much.
The most defensible brand loyalty statistic in 2026 is that repeat buyers spend 69.2 percent more and place 57.6 percent more orders than new buyers, from Bluecore’s analysis of more than 100 retailers. It is transaction data rather than survey data, it holds across every vertical measured, and the sample is large enough to be meaningful. The companion figure from the same source, that 74 percent of customers buy once and never return, is equally well grounded. Both are more useful than the widely quoted claim that a 5 percent retention lift raises profits 25 to 95 percent, which is a 2014 restatement of a 1990 study of three service businesses.
The 25 to 95 percent range is a 2014 Harvard Business Review restatement, and the original 1990 research by Frederick Reichheld and W. Earl Sasser reported 25 to 85 percent. The 1990 study measured three service businesses specifically: an 85 percent profit gain in one bank’s branch system, roughly 50 percent in an insurance brokerage, and 30 percent in an auto service chain. The direction of the finding has held up for thirty-five years and the underlying mechanism is sound. What is not accurate is presenting it as an ecommerce benchmark, since ecommerce did not exist when the research was conducted, or quoting the 95 percent ceiling as though it appeared in the original paper.
A good repeat purchase rate depends far more on your category and purchase cycle than on any cross industry average. Bluecore’s benchmark data puts health and beauty at 21.5 percent, sporting goods and outdoor at 21.2 percent, apparel at 20.2 percent, and the blended average across all verticals at 16.5 percent. Other widely cited sources put the ecommerce average nearer 28 percent, and both can be correct because they measure different store populations over different windows. A supplements brand at 20 percent has a retention problem. A furniture brand at 20 percent is performing well. Benchmark against brands with similar price points and natural repurchase intervals, not against a category average.
AI is shifting loyalty from brands toward outcomes, and 57 percent of consumers now say they would let an AI shopping agent switch brands on their behalf if it found better value. Checkout.com’s 2026 research found that alongside evidence that adoption is moving fast: NIQ reports 42 percent of consumers used an AI tool to shop within the past month. The structural problem for brands is visibility. McKinsey found that only 1 percent of the sources language models cite when answering questions about a brand come from that brand’s own website, meaning your product pages and brand story have little influence on what an agent recommends. Machine readable product and offer data is becoming a retention requirement.
Loyalty programs are worth building, but they are no longer a differentiator, because more than 90 percent of companies now run one. Antavo’s 2026 report found 92.7 percent of program owners saw positive return at an average of 5.3 times, though that figure comes from a vendor survey of the people who purchased the programs and should be read with that in mind. The more actionable finding is that shoppers now belong to about three programs on average and over half would disengage from a program that only rewards purchasing. The sequencing matters most: if your repeat purchase rate is under 20 percent and you have no post purchase email flow, build the flow before the program.
