The 2026 DTC repeat purchase benchmark
- Repeat revenue is inherited, not earned this year. Over the 2025-07 to 2026-08 window, 56.9% of revenue was first orders, 32.9% came from customers acquired before the window opened, and only 10.2% came from repeat orders by customers acquired inside it.
- There is no benchmark repeat rate to hit. Across 15 brands on the same platform, repeat order rate ran from 12% to 67.8%, a 5.6 times spread, median 38.8%.
- Newly acquired customers mostly do not come back, and they decide fast. The median brand saw 5.17% of a cohort order again in month two, 1.49% by month six. 86.7% of a cohort's first 6 months of revenue lands in month one.
- Edition 2 corrects a data defect and withdraws one finding. Two brands' platform-migration imports were being counted as customer acquisition. Removing them cut the sample by a third and dropped the product-category comparison below the brand minimum this series publishes to, so that comparison has been withdrawn rather than restated.
- Average order value predicts nothing. The rank correlation between order value and repeat rate was -0.15 across 15 brands.
This is first-party data. Every figure on this page was computed from order and cohort data in Interconnections' own reporting warehouse, covering 468,386 orders and $69,656,336 of revenue across 17 managed DTC brands between 30 December 2023 and 17 August 2026.
Nothing here is estimated, modelled, surveyed, or drawn from another publisher. No figure was typed by hand: a reproducible pipeline reads the warehouse, computes each number, and this page renders from its output. The method and its limitations are stated in full at the end, including the several things this data cannot answer.
Edition 1 of this report overstated its own sample and has been corrected. Two of the brands in it had migrated their order history onto Shopify, and the import tool stamps every imported order with the date of the import rather than the date of the original sale. Bucketed on that date, one brand appeared to take 131,637 orders in a single month and four the month after.
Counted as acquisition, those imports were 35% of the orders and 56% of the cohort customers behind edition 1's figures. Every sample number on this page has changed as a result: the study covers 468,386 orders rather than 720,140, and $69,656,336 rather than $96,204,898.
The findings themselves held. Every headline figure was reported as a median across brands rather than a pooled total, specifically so that no single brand could drive a result, and those medians are unchanged: the first month still carries 86.7% of a cohort's first 6 months of revenue, and the median brand still sees 5.17% of a cohort return the following month. The pooled figures quoted beside them have moved, and one finding has been withdrawn outright because the correction dropped it below this series' minimum brand count.
A screen now runs before every figure on this page and quarantines any brand-month whose order count dwarfs that brand's own typical month, or whose repeats are already over by day 30. It flags exactly the three months involved here and nothing else.
Why we ran this
Every growth plan has the same paragraph in it. Acquisition costs more than it used to, so the margin has to come from lifetime value: buy the customer once, earn on the second, third and fourth order. The paragraph is usually followed by a repeat rate target borrowed from an industry report.
We had the order history sitting in our reporting warehouse, so we looked at what actually happened. 468,386 orders, $69,656,336 of revenue, 17 DTC brands, 96,631 customers tracked from their first order forward.
Two things came out of it that changed how we plan. The first is that the repeat revenue showing up in a profit and loss statement this year was mostly produced by acquisition done in previous years, which makes it a much slower lever than it is usually sold as. The second is that the spread between brands is so wide that any single benchmark number is meaningless, and the obvious explanation for that spread, that some categories just repeat and others do not, turns out to explain far less of it than we expected.
most of the money is a first order
Start with the simplest question anyone can ask of a store: over some period, where did the revenue come from? Not which channel, but which customers, and when those customers were first acquired. We split every dollar across 16 brands over 2025-07 to 2026-08 into three buckets: first orders from customers acquired during the window, repeat orders from those same customers, and orders from customers who were already on the file when the window opened.
Two thirds of it, 56.9%, is somebody's first order. That part is unsurprising. The interesting part is the other third. Of it, 32.9 percentage points came from customers who were already there before the window started, and only 10.2 percentage points came from customers acquired inside the window coming back for more.
That ratio is not a quirk of one brand. Taking each brand on its own and then reading the middle of the distribution, the median brand drew 29.6% of its revenue from pre-window customers and 9.7% from in-window repeats. The pre-window share ran from 0% to 82.9%, the low end being brands that simply have not been trading long enough to have accumulated a base.
The practical consequence is a timing problem rather than a strategy problem. Retention work compounds, which is exactly why it pays out slowly. A brand that fixes its post-purchase flows in January is not moving this year's repeat line very much, because this year's repeat line is being produced by customers acquired in the two years before that. It is moving the line two years out. That is a good reason to start, and a bad reason to promise a payback inside the year.
there is no repeat rate you are supposed to hit
The store-level question is easier to measure than cohorts and it is the number most operators actually quote: of all the orders placed, what share came from someone who had bought before? Across 15 brands with at least a year of trading and at least 500 orders each, the pooled answer is 40.9% of orders and 41.1% of revenue.
Quoting either of those as a benchmark would be close to useless, because of how they are distributed.
The lowest brand sits at 12% and the highest at 67.8%. That is a 5.6 times spread between two businesses selling physical products to consumers on the same platform. Half of them fall between 20.6% and 52.3%, which is itself a range wide enough that a brand at either end would draw completely different conclusions about whether it has a retention problem.
Revenue share is wider still, from 9.6% to 85.7%, median 43.1%. Repeat revenue share exceeds repeat order share at almost every brand, which tells you returning customers spend more per order than new ones. That is one of the few things in this study that behaves the way the textbooks say it should.
a new customer decides quickly, and mostly decides no
Store-level repeat rate mixes together customers acquired last week and customers acquired three years ago. To see how a customer actually behaves after their first order, you have to follow cohorts: group customers by the month they first bought, then watch that group forward.
This is where the analysis is most vulnerable to getting it wrong, so the cohort is restricted. A group acquired last month has had one month to come back; including it alongside a group acquired a year ago would make early months look artificially dominant. Only cohorts observed for a full 6 months are used: 133 cohorts across 17 brands and 96,631 customers, with no single brand contributing more than 20% of them.
For the median brand, 5.17% of a cohort places an order in the month after they arrive. By the fifth month it is 1.49%, and the curve has been close to flat for three months by then. The shape matters as much as the level: this is not a gentle decline that you can arrest with a well-timed campaign in month four. Almost all of the second-order behaviour that is going to happen happens immediately, and then the cohort goes quiet.
Reading it in revenue rather than customers gives the sharpest version of the same fact. For the median brand, 86.7% of everything a cohort will spend in its first 6 months is spent in the first month. Pooled across all brands it is 82.4%. Half the brands sit between 77.2% and 90.1%.
The range runs from 29.3% to 95.6%. At the top end are brands where the first order is effectively the entire relationship. At the bottom is a single brand where most of the six-month value arrives after the first order, which is what a genuine subscription dynamic looks like and which almost nobody in this sample has.
What we can no longer say about product category
Edition 1 of this report carried a fourth finding: that repeat rate is not simply a function of what a brand sells. It compared brands whose product is consumed and rebought against brands selling durable or gift purchases, and reported that category shifts the median without deciding the outcome.
That finding is withdrawn. Correcting the migration defect described above removed two brands' import months, and both of those brands then fell under the 365-day trading minimum this study requires before a repeat rate means anything. The replenishable group went from six brands to 4, below the 5-brand floor Interconnections publishes to.
Dropping the tenure threshold by four days would restore the group to five and the finding with it. That is precisely the post-hoc tuning this methodology exists to prevent, so it has not been done. The comparison will return in a later edition when the affected brands have traded long enough to qualify on their own.
What survives from it, because it never depended on the category split: average order value has no relationship with repeat rate. The rank correlation across 15 brands is -0.15. Discount depth correlates at -0.3, which points against the idea that discounting builds a repeat habit and is too weak at this sample size to claim either way.
What you have been told, and what the data shows
| The common claim | What this data shows |
|---|---|
| You can afford a higher acquisition cost because of lifetime value | For the median brand, 86.7% of a cohort's first 6 months of revenue arrives in the first month. The later value exists but it is small and slow. |
| Aim for the industry benchmark repeat rate | The range across 15 brands is 12% to 67.8%. No single figure describes it. |
| Retention is cheaper than acquisition, so shift budget there | Only 10.2% of revenue came from in-window repeat orders. Shifting budget out of acquisition shrinks the base that produces repeat revenue later. |
| Platform data is clean enough to analyse as it arrives | Two of these 17 brands had migrated onto the platform, and the import stamped a decade of orders with the import date. Unscreened, that was 35% of the order base. |
| Discounting trains customers to come back | Correlation of -0.3, pointing the other way, and too weak to claim. |
What is actually happening underneath these numbers
A repeat base is a stock, and acquisition is the flow that fills it. That framing explains most of what is above. The stock at any moment is everybody you have ever acquired who is still willing to buy. This year's repeat revenue is drawn from that stock, which was filled in previous years. This year's acquisition adds to the stock, and most of the return on that addition shows up later.
It also explains why the spread between brands is so wide and so weakly related to category. Two brands selling the same kind of product will have very different stocks depending on how long they have been trading, how fast they have been acquiring, and how much of what they acquired was worth keeping. A brand that has been buying customers hard for four years has a large stock and will report a high repeat rate even if its post-purchase experience is mediocre. A brand growing quickly from a standing start will report a low one even if its product is excellent, because the denominator is full of customers who only just arrived.
That last point is worth sitting with, because it inverts how the metric is usually read. A falling repeat rate can be a symptom of successful acquisition. If you double the rate at which you bring in new customers, the share of orders coming from returning ones goes down mechanically, and nothing about your retention has got worse. Read on its own, the number will tell a growing brand that it is failing at exactly the moment it is succeeding.
What to watch, and what it actually means
The practical version of everything above, as a lookup.
What we would actually do with this
Judge retention on cohorts, not on the blended rate. The blended repeat rate moves with acquisition volume and with how long you have been trading, neither of which is retention. The question that survives is whether the month-two share for the cohort you acquired this quarter is better than the one you acquired last quarter.
Spend the retention effort inside the first 30 days. For the median brand in this sample, 86.7% of six-month cohort revenue lands in the first month and the repeat curve is nearly flat from month three. Effort aimed later is aimed at people who have already stopped paying attention.
Underwrite acquisition on the first order. Given how much of early cohort value is the first order, a payback model that leans on months two through six is leaning on 13.3% of the money for the median brand. Interconnections plans paid media to a contribution target on first-order economics and treats later revenue as upside rather than as the thing that makes the maths work.
Stop benchmarking against published averages. With a 5.6 times spread between comparable brands, an external benchmark cannot tell you whether you have a problem. Your own trend can.
How we would structure retention work given this
Nothing here argues against retention work. It argues against expecting it to behave like a performance channel, with spend in one month and a measurable return the next.
The sequencing that follows from the data is roughly this. First, treat the first 30 days as the whole battleground, because that is where the behaviour actually happens. Second, measure by cohort so that improvements are visible before they are large enough to move the blended number, which on these figures could take a year or more. Third, keep funding acquisition, because the stock that produces future repeat revenue is filled by acquisition and nothing else fills it.
The uncomfortable version, which we would rather state than dress up: for a brand with a young customer file, there is no amount of retention work that fixes this year. The lever exists and it is worth pulling, and the return arrives on a schedule most quarterly plans are not written to accommodate.
Frequently asked questions
What is a good repeat purchase rate for a DTC brand?
There is no single answer, and that is the finding rather than a dodge. Across the 15 brands Interconnections measured, repeat order rate ranged from 12% to 67.8%, a 5.6 times spread, with a median of 38.8%. Half the brands sat between 20.6% and 52.3%. A number quoted as the DTC average is describing one brand's business model, not a target you should be held to.
How long does it take a new ecommerce customer to buy again?
Mostly they do not. In the 6-month cohorts Interconnections analysed, the median brand saw 5.17% of a new cohort place a second order in the following month, falling to 1.49% by month five. Across the whole horizon, 86.7% of a cohort's revenue arrived in the first month for the median brand.
Is repeat revenue worth investing in if most customers never return?
Yes, but not on this quarter's numbers. Interconnections found that 32.9% of revenue in the window came from customers acquired before it opened, against 10.2% from repeat orders by customers acquired inside it. Repeat revenue is real and it compounds, but it is produced by acquisition done in earlier years. Treating it as a lever you can pull this quarter is the mistake.
Does repeat purchase rate depend on what you sell?
Interconnections compared brands whose product is consumed and rebought against brands selling durable or gift purchases, and edition 1 of this report published the result. That comparison is withdrawn in edition 2: correcting a data defect removed two brands from the replenishable group, leaving 4, below the 5-brand minimum this series publishes to. What does survive is that average order value has no relationship with repeat rate at all, a rank correlation of -0.15 across 15 brands, and that the range between the highest and lowest brand is 5.6 times regardless of what they sell.
Does a higher average order value mean worse retention?
No relationship at all in this data. Interconnections measured the rank correlation between average order value and repeat order rate across 15 brands at -0.15, which is indistinguishable from zero. Expensive brands and cheap brands are found at both ends of the retention range.
Does discounting buy loyalty?
This data points gently the other way, but not strongly enough to claim it. The rank correlation between discount depth and repeat order rate was -0.3 across 15 brands. That is the opposite sign to the theory that discounting builds a repeat habit, but at this sample size Interconnections treats it as unproven rather than as a finding.
What we would measure next
Three things would sharpen this materially. The first is the interval between orders in days rather than months, which would replace the coarse monthly curve with an actual distribution and show whether the replenishment cycle is visible in the data. The second is entry product: whether the item somebody buys first predicts whether they come back, which given how decisive the first order turns out to be may be the highest-leverage question in the whole subject. The third is a longer cohort horizon, which needs time to pass rather than any new analysis.
What this study cannot tell you
It cannot rank categories, and in this edition it cannot compare them at all. The sample holds one supplement brand and no skincare, coffee, or pet consumables. After the migration correction the replenishable group holds 4 brands, below the 5-brand floor, so the comparison published in edition 1 has been withdrawn.
It says nothing beyond 6 months. Only a small number of cohorts have been observed for a full year, and one brand contributes the large majority of the customers in them. Publishing a twelve-month figure would effectively be publishing that one advertiser's economics, so no twelve-month claim is made anywhere on this page.
Discounting is unproven here. The correlation of -0.3 points against the loyalty-through-discounting theory, and at 15 brands that is suggestive at best. We have deliberately kept it out of the findings.
Guest checkouts are excluded. Cohort analysis needs a customer identity to follow. Orders placed without one cannot be assigned to a cohort, so a brand with heavy guest checkout will have some genuine repeat behaviour that is invisible here.
It is not causal. Nothing on this page establishes why one brand retains better than another. It establishes how much they differ, and rules out three of the explanations most commonly offered.
Method
Order-level and cohort data for 17 managed DTC brands, read from Interconnections' reporting warehouse. Two views are used and they answer different questions. The store-level view counts every order and asks what share came from a returning customer. The cohort view groups customers by the month of their first order and follows each group forward.
This is edition 2, covering 30 December 2023 to 17 August 2026. The window is frozen against a fixed snapshot rather than read live, because the underlying tables are rebuilt nightly and the cohort table is fully overwritten on every run. A report wired straight to them would quietly restate its own findings every time it was rendered. Future editions extend the end date; the start date does not move and published figures are not revised.
Censoring is controlled in both directions, because it is the failure mode that would invalidate everything else. Cohorts are only included once they have been observed for a full 6 months, so a recently acquired group cannot inflate the apparent dominance of the first month. Brands are only included in store-level figures once they have at least 365 days of trading and 500 orders, so a recently onboarded brand cannot depress the repeat rate simply by not having existed long enough.
Every figure is reported as a per-brand median alongside the pooled value. Where the two disagree the pooled figure is being driven by one large brand, and the median is what gets published. No individual brand is named and no per-brand row appears anywhere on this page. Every published figure aggregates at least 5 brands. Average order value and discount depth are never shown per brand, since either would identify an advertiser; only rank correlations across the whole sample are published, and a rank correlation reveals nothing about any single brand.
- Repeat order rate is orders from returning customers divided by all orders, computed on pooled sums per brand rather than as an average of daily ratios.
- Cohort month zero is the calendar month of the customer's first order, so a fast second purchase inside those weeks falls into month zero rather than month one.
- Revenue figures are order totals as recorded by the store, including shipping and tax and net of refunds where the platform reports them that way.
- The replenishable classification is ours, made from each brand's top-selling products rather than from any self-description, and is stated as a judgment rather than a measurement.
Because most of a customer's early value is the first order, Interconnections underwrites acquisition on first-order economics rather than on projected lifetime value. The service behind that is paid media management, the post-purchase work is email and retention marketing, and the first-order conversion work is conversion rate optimization. The companion study on creative is the Meta creative fatigue benchmark, what this produced is in the case studies, and a growth diagnostic is how an engagement usually starts.