How to Reduce Returns in Fashion Ecommerce: The Lever Most Brands Are Missing

September 3, 2026 | For Brands

Returns are the tax every fashion brand pays for selling online, and it is the only line on the P&L that grows in direct proportion to how well marketing performs. Sell more, and you return more. The industry has spent the better part of a decade building tools to bring that number down — size recommenders, virtual try-on, 3D body scanning, fit finders trained on millions of garment measurements. All of them attack the same half of the problem, and they attack it well.

This piece is about the other half: the returns that happen even when the size was right. That bucket is bigger than most brands assume, almost nothing on the market addresses it, and the reason is structural — the missing information was never about your garment. It was about the wardrobe your garment was going into.

The Numbers Every Fashion Brand Already Knows

The headline figures are not in dispute. According to the 2025 Retail Returns Landscape, published by the National Retail Federation with Happy Returns, US consumers returned $849.9 billion of merchandise in 2025 — 15.8% of annual retail sales. Online, that rate climbs to 19.3%. Industry benchmarks consistently put apparel at the top of every category table, with return rates around 25% and footwear higher still.

The direct cost is worse than the rate implies, because a returned garment is not a neutral event that simply reverses a sale. Estimates of all-in processing — inbound shipping, receiving, inspection, grading, repackaging, re-ticketing, restocking — commonly land somewhere between $10 and $65 per item, and a meaningful share of returned apparel never goes back on the shelf at full price at all. It gets marked down, liquidated, donated or written off.

Then there is the part that rarely gets modelled properly: what returns do to acquisition economics. The arithmetic is simple and unforgiving. If you are acquiring a customer at $75 and 26% of what they order comes back, your effective cost per kept order is not $75. It is $75 ÷ 0.74, or about $101. Every point of return rate silently re-prices your entire media budget, and it does so in a market where apparel acquisition costs have been climbing for years.

And the damage does not stop at the refund. The same NRF research found 71% of consumers are less likely to shop with a retailer again after a poor returns experience, and 82% now name free returns as a major consideration when deciding where to buy. You are paying for the return twice: once in logistics, and again in the retention you lose when the experience is bad — or in the margin you give away keeping it good.

Fit Technology Solves 39% of the Problem

Here is where the conversation usually stops, and it should not. The most useful public dataset on why apparel comes back is PowerReviews' Apparel Shopping Trends research — a survey of 25,452 US consumers. Over half of them, 53%, had returned an apparel purchase in the previous 90 days. Asked why:

• It did not fit — 39%
• It did not look the way they thought it would — 28%
• They bought multiple sizes and sent back the rest — 13%
• It did not match the description — 13%
• Damaged or defective — 10%

Fit is the largest single reason, and the industry has responded rationally. Sizing is fundamentally a measurement problem: there is a correct answer, the shopper does not know it, and technology can supply it. Size recommenders, fit prediction, AR try-on and body scanning all exist to close that gap, and where they are implemented well they work.

Now look at the second row. Twenty-eight percent of apparel returns happen because the item did not look the way the shopper expected — and that is not a measurement problem. There is no correct answer to supply. Nobody sent the wrong garment. The photography was accurate, the size was right, the description was fair. The shopper simply discovered something after delivery that they could not have discovered before it.

No size chart fixes that row. Neither does better product photography, or another angle on the PDP, or a more accurate swatch. Brands have poured budget into all three, and the 28% has not moved, because the missing information was never held by the retailer in the first place.

The Missing Information Is the Customer's Existing Wardrobe

Almost nobody buys clothes in isolation. A shopper looking at a jacket is not really asking "is this a good jacket." They are asking a much harder question, usually without articulating it: does this work with what I already own?

That question is a compatibility bet, and until it is settled the purchase is provisional. The shopper places the bet at checkout with almost no information, resolves it in their bedroom against a real wardrobe, and if the answer is no, the garment goes back — filed under "did not look the way I thought it would," because that is the only box on the form that comes close.

The reason this has gone unaddressed is not that it is unimportant. It is that the data required to answer it has never been available to the seller. A retailer knows its own catalogue in enormous detail and knows essentially nothing about the forty other garments the shopper will style the purchase against. The one party who could answer the question — the shopper — cannot do it either, because they are standing in a shop or scrolling a PDP, and their wardrobe is at home.

Nouva is built on the one dataset that closes that gap: users photograph what they already own, and the app works from that wardrobe. Everything else follows from it. When Nouva generates a complete outfit, it is assembling pieces the user actually owns and scoring the combination for colour harmony. When it recommends something to buy, it is doing so against a specific, named gap in a specific, real wardrobe.

What a Catalogue Partnership Actually Does for Your Brand

This is the part that matters commercially, so let us be concrete about what changes when your products are in that catalogue.

Placement is gap-driven, not impression-driven. In Nouva's Stylist tool, the user's own closet is searched first, and a product is only suggested when their wardrobe genuinely cannot fill a slot in the look they are building. Your item does not surface because a shopper matched a demographic segment or a lookalike audience. It surfaces because the outfit in front of them is missing exactly that piece and they do not own it. That is about as far from a spray-and-pray impression as retail media gets.

The product arrives inside an outfit, not a grid. A shopper who sees your jacket in a Nouva result is not evaluating it against a white background. They are seeing it placed in a complete look built from clothes they already own — trousers they wear, shoes they like, in colours that have been checked against each other. The "what would I wear this with" question, which is the question that produces the 28%, is answered before checkout rather than discovered after delivery.

Pre-purchase filtering works in your store, too. ShopScan lets a shopper point their phone at an item on the rail — no barcode, no integration — and get a verdict against their own wardrobe: whether they already own something close to it, and which of their pieces it would actually go with. That is a decision aid at exactly the moment the compatibility bet is being placed.

Style and colour matching is the ranking, not a filter. Your catalogue metadata — colour, category, material, occasion, style — is read and used to place products, not just to populate a filter sidebar. Products that genuinely complement a user's wardrobe surface; products that do not, do not.

Stock discipline is enforced. Availability is resolved before anything is displayed. A shopper is not sent to a sold-out PDP, which is a small thing that quietly destroys trust in every affiliate surface that ignores it.

Attribution carries context. Clicks are attributed with the outfit they came from, so the conversation about performance can be about which styling contexts sell your product, not just how many clicks a placement produced.

Why "Tell Them Not to Buy It" Protects Your Margin

The first objection brands raise is a fair one. ShopScan will sometimes tell a shopper to skip an item — including yours — because they already own something close to it. Why would a brand pay to be in a system that does that?

Because the alternative is not a kept sale. It is a return.

Run the two outcomes side by side. Take a £90 garment at a 55% gross margin, so £49.50 of gross profit if the order sticks.

If the sale is kept, you recognise £90, bank £49.50 of gross profit, earn back the acquisition cost, and keep a customer who is more likely to come back.

If the sale comes back, every one of those reverses. Revenue goes to zero and so does the gross profit. The outbound shipping is already spent and unrecoverable. You now add £10 to £30 of all-in return processing on top. The garment re-enters inventory to be resold at a markdown or written off entirely. The acquisition cost was still paid, with nothing to show for it. And the customer relationship is measurably worse, not neutral — this is the same experience that made 71% of shoppers in the NRF research less likely to buy from a retailer again.

The suppressed sale is not a lost £49.50. It is a £10 to £30 cost avoided, an acquisition spend redirected to a shopper who will keep the item, and a customer who now trusts the recommendation the next time it says buy. A recommendation engine that never says no is not a recommendation engine — it is a banner, and shoppers price it accordingly.

How Spend Actually Works Here

Every partner asks the same question, and it deserves a direct answer: what does paying more get me?

Relevance is a gate. Budget is a dial inside that gate.

If a garment does not work with a shopper's wardrobe, no level of spend puts it in front of them. That door does not open, at any price. But among the shoppers a product genuinely suits — the ones whose wardrobe has exactly the gap it fills — a larger commitment means more of them see it, more often, and in more of the styling contexts where it fits. Paid placement is disclosed to users.

That is a real, buyable lever, and it is worth being clear about why it is drawn in this particular place. In a conventional ad auction, budget buys reach, and relevance is a targeting preference you can loosen when you want more volume. Here it is the other way round: relevance is fixed and reach is what you buy. Loosening the gate would be the easiest thing in the world to sell you and the fastest way to destroy what you are actually buying — because the moment products start appearing in front of people they do not suit, the returns come back, the recommendations stop being trusted, and the click you paid for is worth what a banner impression is worth.

So the honest summary is this: you are not buying impressions, and you are not buying a better styling verdict. You are buying more of the demand that already fits you — and that pool is finite, which is precisely why it converts.

What This Does Not Do

Three honest limits, because a pitch without them is not worth reading.

It does not fix fit. The 39% is a real problem and this is not a solution to it. If you do not already have size guidance on your PDPs, that is the higher-priority investment. Wardrobe-context matching is complementary to fit technology, not a substitute — the two address different rows of the same table.

It does not fix quality or description gaps. The 13% who returned because the item did not match the description, and the 10% who received something damaged, are merchandising and operations problems. No styling layer touches them.

We are not publishing a return-rate figure for partners, because we do not have an honest one yet. Nouva is a growing app, not a mass channel, and the return-reduction argument above is a mechanism argument grounded in third-party data about why apparel comes back — not a measured result from our own partners. Any vendor quoting you a precise percentage lift for your specific catalogue is guessing. What we can commit to is measuring it properly with the brands who come in early, and sharing what the measurement says.

What a Partnership Requires From You

Practically, a catalogue needs to arrive with enough structure to be styled against:

• Product imagery on a clean, consistent background. The same photography principles that make a wardrobe app work apply to catalogue imagery — accurate colour matters more here than production polish, because colour is doing real work in the matching.
• Category and garment type, using terms a stylist would recognise.
• Colour, as specifically as you hold it.
• Material and composition, which drive seasonal and weather-appropriate placement.
• Occasion or style descriptors, if you hold them.
• Price, currency and live stock status.
• Deep links to the product page, with whatever tracking parameters your programme uses.

Nouva currently serves two affiliate catalogues — US retailers in USD, and UK retailers in GBP for the rest of the world — so those are the two markets a catalogue can be listed into today. If your feed already runs through an affiliate network, it very likely contains most of the above already.

Frequently Asked Questions

Is this an ad network? Not in the usual sense. You are not bidding for impressions against a broad audience, and no level of spend puts a product in front of someone it does not suit. What a larger commitment does buy is reach within the shoppers it does suit — more of them, more often, across more styling contexts. Relevance is the gate; budget is the dial inside it. Paid placement is disclosed to users.

So can we pay to appear more? Yes — to more of the right people. That is the whole distinction. Spend scales how much of the matched demand you capture; it does not widen the definition of who is matched, because widening it is what would put the return rate back up and make the placement worthless to both of us.

How is this different from a size recommender? A size recommender answers "which size of this garment fits your body." Nouva answers "does this garment work with the clothes already in your wardrobe." Different questions, different return reasons, and they compose well — most brands should have both.

What does it cost? There is no public rate card. Terms depend on catalogue size, market, the level of reach you want inside the matched audience, and how you already handle affiliate distribution — worked out directly rather than through a self-serve tier.

Which markets can we list in? US (USD) and UK (GBP) today. UK is the default catalogue for shoppers outside the US.

Do you need our customers' data? No. The wardrobe data belongs to the Nouva user and is not shared with brands. What you receive is performance reporting on your own products.

What happens when an item goes out of stock? It stops being shown. Availability is resolved before display, so shoppers are not sent to dead product pages.

How would we measure whether it works? Click-through and conversion by styling context on our side, and return rate on the cohort attributed to Nouva on yours. That second number is the one that matters, and it requires your data to compute — which is why it is a joint measurement rather than a dashboard we hand over.

The Short Version

Every fashion brand is already spending heavily on the 39% of returns caused by fit, and should keep doing so. Almost nobody is spending anything on the 28% caused by a garment not looking the way the shopper expected — not because it is unimportant, but because the information needed to prevent it lives in the customer's wardrobe, where no retailer can see it.

Nouva can see it, because users put it there. Products placed against that context are not being shown to a segment. They are being shown to a person whose specific wardrobe has a specific gap that your product closes — and who has been told, before paying, exactly what they would wear it with.

If you run a fashion brand or a catalogue and that is the kind of demand you would rather buy, get in touch — partnership enquiries go straight to the team that builds the product, and a person replies.

---

Nouva is an AI stylist app for iOS and Android. It builds outfits from clothes people already own, scores them for colour harmony, and checks an outfit before you leave the house. Brand and catalogue enquiries: nouva@uniquepresident.com.