A return does not feel like a marketing problem. It arrives as a padded envelope on a Tuesday, someone opens it, checks the item, puts it back into stock if it can go back into stock, and issues the refund. It is handled by whoever handles shipping. It gets recorded, if it gets recorded at all, as a cost of doing business.
That is the mistake, and it is an expensive one. Every item in that pile was a sale you already paid to get. You bought the click. You paid the transaction fee. You paid to put the thing in a box and move it across the country. Then you paid again to bring it back, and depending on what it is, you may not be able to sell it a second time at full price.
The National Retail Federation's 2025 returns research puts the online return rate at 19.3%, against 15.8% across retail as a whole. Those figures come from large merchants and yours will differ, but the shape of the thing is hard to argue with: online, close to one order in five comes back, and it comes back at roughly a quarter more often than it would from a shop floor.
And here is the part almost nobody acts on. The decision that produced that return was made days earlier, on your product page, by what you led somebody to expect.
Your Return Rate Is Not a Logistics Number
Returns get filed under operations because that is where they are processed. That is not where they are caused.
By the time an item is in the returns pile, every decision that mattered has already been made. Someone looked at five photographs and a paragraph of copy and formed a picture in their head of what was going to arrive. The box turned up. The picture did not match. Everything after that point is administration.
Which means the returns pile is not really telling you about your warehouse. It is telling you about your merchandising, your photography, your product copy, and increasingly, your advertising. Those are all things you can change. The padded envelope on a Tuesday is not.
This reframe matters because it changes who is responsible for the number. When returns belong to operations, the fix is always cheaper packaging, a stricter policy, or a better rate with the carrier. Those are real levers and they are small ones. When returns belong to merchandising, the fixes get much larger, because you are working on the cause instead of the consequence.
There is a second reason to care, and we have written about it before in the context of working out which products actually earn. Returns do not spread themselves politely across a catalog. They cluster. Anything involving fit, color or feel takes them disproportionately, which means one or two products are usually carrying most of your return cost while everything else looks clean. Averages hide exactly the product you need to find.
"Too Big" and "Too Small" Are Not the Same Problem, and Your Data Cannot Tell Them Apart
This is the cheapest fix in this article and most stores skip it.
When somebody starts a return, you get one chance to ask why. Most stores either do not ask, or they ask with a free-text box. A free-text box feels generous and it is close to useless, because nobody can count it. Three hundred returns produce three hundred slightly different sentences, and the answer to "why are people sending this back" stays permanently out of reach.
Fix the question, not the answer. Give six or seven fixed reasons that match what you actually sell, plus a single "other" with a text field for the genuinely unpredictable. For apparel that is: too small, too big, not the color I expected, quality was not what I expected, arrived damaged, arrived too late, changed my mind. For furniture or hardware it is a different seven. The point is that they are countable.
Then read them monthly and do something. Daniel Chabert, who runs a conversion agency and has owned stores himself, makes the point that plenty of businesses collect this data and never look at it, and looking at it is where the money is.
The reason this matters so much is that "too small" and "too big" are opposite problems with opposite fixes, and a raw return rate averages them into a single meaningless number. If a product comes back because it runs small, you adjust the sizing guidance and you tell people to size up. If the same product comes back because it runs large, the fix is the exact reverse. A store looking only at "this product has a 30% return rate" cannot tell those apart, so it does nothing, or it does the wrong thing.
Two more reason codes worth separating, because they are not merchandising problems at all: arrived damaged points at packaging, and arrived too late points at fulfillment. Both are worth knowing. Neither should be mixed in with fit.
The Bracketing Problem You Paid to Create
The NRF found that nearly two-thirds of consumers admit to at least one costly returns behavior. The one you are most likely funding is bracketing: ordering the same item in three sizes, keeping one, sending two back.
It is tempting to read that as customers behaving badly. It is more useful to read it as customers behaving rationally. If your page cannot answer "which size am I", and returns are free, then ordering three and deciding at home is the only reliable way to get the right one. You built that experiment and you are paying for it.
Which means the lever is not a stricter returns policy. It is confidence at the point of choosing.
Three things move it, in rough order of effort:
Show the model's measurements and the size they are wearing. Not a size chart somewhere else on the page. A small note right in the gallery: this model is 5'9" and wearing a medium. It costs nothing, it happens at the exact moment of doubt, and it lets someone compare against a body instead of a table of numbers.
Replace the static size chart with something that answers a question. A chart asks the customer to measure themselves accurately and interpret the result. Most people will not. A guided fit tool asks height, weight, usual size elsewhere and fit preference, and returns an answer. Chabert's point is that this raises conversion at the same time as it lowers returns, which is unusual, because most returns tactics trade one against the other.
Make your reviews answer the fit question specifically. A general five-star rating does not help someone choose a size. A review that rates true-to-size, comfort and durability separately, and surfaces "runs small" as a summary line, does. We have made the broader case for reviews as the thing strangers actually believe, and this is the version of it that pays twice.
None of this is limited to apparel. Anything where fit, scale or compatibility is uncertain has the same problem wearing different clothes. Furniture that turns out bigger than the room. A part that does not fit the model year. A shade of paint that looked different on a phone screen.
Your Photographs Make a Promise the Product Has to Keep
The single most common thing a customer says when an item goes back is that it was not what they expected. That is a photography brief, not a complaint.
Three things routinely cause it, and all three are fixable in an afternoon of shooting:
Scale. A product photographed alone on a white background has no size. It could be six inches or two feet. Put a hand in the frame, or a room, or something everyone recognizes. This one causes a startling number of furniture and homeware returns.
Color. Screens vary and there is no fixing that entirely, but shooting under consistent, neutral light and saying so in the copy gets you most of the way. If a fabric reads warmer in daylight than under a lamp, write that down. Naming the variance is more convincing than pretending it does not exist.
Texture and weight. People form expectations about how substantial something will feel, and photographs are bad at conveying it. Copy is good at it. Say the weight. Say the thickness. Say what it feels closest to.
The instinct in ecommerce is to make everything look as good as possible, and that instinct quietly generates returns. The photograph that oversells converts better and comes back more often, and you pay for the round trip both ways. The goal is accuracy that flatters, not flattery that misleads.
The Return Rate That Belongs in Your Ad Account
Here is the part that almost nobody runs, and it is the reason we care about returns at an advertising agency at all.
Your return rate is not one number. It is an average of very different numbers hiding underneath it, and one of the ways it splits is by where the customer came from.
Segment your returns by acquisition channel. Then by campaign. Then, if the volume supports it, by audience or ad. What you are looking for is any source whose return rate runs meaningfully above your store average.
When you find one, you have found an ad that is over-promising. That is the whole diagnosis. A campaign bringing in customers who send the goods back at twice the normal rate is not a targeting problem or a landing page problem. It is a promise problem: the creative set an expectation the product does not meet, and the return is the customer telling you so, several weeks later, in the most expensive way available.
This matters more than it sounds, because of how ad platforms report. Your ROAS is calculated on gross revenue. The platform does not know about the refund. So a campaign can look like your best performer in the ads dashboard and be your worst performer in the bank, and nothing in the reporting will ever surface the gap. You have to go and look.
How to actually do it, without buying anything:
Pull your refunds for the last ninety days with the order ID attached. Ninety, not thirty, so one promotion does not distort it. Match those order IDs against the source data you already have, whether that is UTM parameters on the original order, your platform's own sales-by-channel report, or the order attribution your store records at checkout. Group by channel and campaign, divide returns by orders in each group, and compare each one against the store average.
Then go and look at the outliers. In our experience the ads that come back highest are usually the ones leaning hardest on a single flattering image, a heavy discount, or an aggressive claim about fit or finish. Discount-led campaigns are worth checking specifically, because a price cut widens the audience to people who were only ever there for the price, and that cohort behaves differently after the box arrives.
What you change is the ad, not the warehouse. Rewrite the claim. Swap the hero image for one that shows scale honestly. Put the fit guidance in the ad copy instead of hoping the product page catches it. If you are running Shopping campaigns, the same logic applies to the product feed, which is doing the work your keywords used to do - the title and image in the feed are setting the expectation before anyone reaches your site at all.
This is also the number that should change how you judge whoever is running your ads. If nobody has ever shown you return rate by campaign, nobody is optimizing against what you actually keep. It is one of the first things we look at on an ecommerce account, and it is the kind of thing that is easy to see from outside and almost invisible from inside a dashboard. If you would rather have someone handle that side of the store for you, that is what we do; if you would rather run it yourself, the query above is genuinely all it takes to start.
An Exchange Is a Recovered Sale. A Refund Is a Lost One.
Everything above is about preventing the return. This section is about what happens once somebody has decided.
Most return flows are built to process a refund efficiently. That is the wrong goal. The moment somebody starts a return, you still have an interested customer, a product they wanted enough to buy, and a problem that is usually specific and solvable. A refund throws all three away. An exchange keeps them.
Very few small stores offer an exchange as the first option, and it is mostly an accident of tooling: the returns portal was set up to issue refunds, so that is what it does. Reversing the order of those two options is a configuration change that recovers real revenue.
Make the exchange the path of least resistance. When somebody selects "too small", the next screen should offer the next size up, already selected, with free shipping on the replacement. Not a form. Not a store credit they have to remember to use. The thing they wanted, in the size that fits.
Chabert's sharpest operational suggestion is to put live chat on the returns page itself rather than the product page. It is a strange idea until you think about who is on that page: someone actively deciding whether you keep the money. A short conversation at that exact moment, offering an exchange or an alternative, converts refunds into sales at a rate that is easy to measure and easy to justify.
And if the exchange is not possible, store credit at a small premium over the refund value is worth offering. Some people will take it. Every one who does is a customer you keep rather than a customer you refund, which connects directly to why the second order is where the profit lives.
What You Say Before the Box Arrives
There is a window between the order and the delivery where the customer is forming an opinion with no product in front of them, and most stores use it to send a shipping notification and nothing else.
Two things belong in that window.
The first is anything that pre-empts a surprise. Care instructions. What to expect on first use. If the fabric relaxes after a wash, say so before the wash, not in an FAQ nobody finds. If the color deepens, or the finish settles, or the first few uses feel stiff, say it. A surprise that was predicted is a feature. The same surprise unannounced is a defect. This costs one email and it is the same window your post-purchase flows are already using.
The second is speed. Waiting builds regret. Somebody who ordered on impulse and then waited nine days has had nine days to reconsider, read reviews of your competitors, and cool off. By the time the box lands, the purchase belongs to a version of them that no longer exists. Faster shipping is a returns lever as much as a conversion lever, which is why it is often worth offering a cheap expedited option rather than only free-and-slow.
Some Returns You Should Stop Trying to Prevent
It would be easy to read all of the above as a war on returns. It should not be.
A generous returns policy sells more than a restrictive one. That is the uncomfortable arithmetic at the center of this subject: the same policy that produces returns is also producing sales that would not otherwise happen, because it removes the risk of buying something you cannot touch from a store you have not used before. Tighten it too far and you will save on reverse shipping while quietly losing more at checkout, where you will never see the loss. We have written about how much of that friction lives in the checkout itself.
So the target is not zero. The target is fewer returns caused by inaccuracy, with the returns caused by generosity left alone.
Three places where the honest answer is to stop optimizing:
Final sale, used deliberately. A clearly marked final-sale category, with a real discount attached, is fair and most customers accept it. It works for discontinued stock and end-of-line items. It does not work as your default policy, and applying it broadly reads as a store protecting itself from its own products.
Serial returners. A small number of customers return most of what they order, and some of them are wearing it first. You can identify them by return rate and handle them specifically, up to and including asking them to pay return shipping, as long as your policy says you reserve the right to. What you should not do is punish everybody for the behavior of a few, which is what a blanket policy change does.
Products with a structurally high rate. Some categories come back far more than others no matter how well you sell them, and fitted clothing is the obvious one. If the product is still profitable once the return cost is subtracted, a high return rate is not a problem, it is a characteristic of what you sell. Work out the contribution per unit after returns and judge the product on that number rather than on the rate itself.
Where to Start This Week
Four things, in order, and none of them require new software.
Turn on structured return reasons. Six or seven fixed options plus one "other". If your returns portal cannot do this, your ecommerce platform's own returns feature almost certainly can.
Give it a month, then read what it says. Sort by product. You are looking for the one or two items generating a disproportionate share, and the specific reason attached to them.
Pull returns by campaign for the last ninety days. Compare each source against your store average. Anything meaningfully above it is an ad making a promise your product is not keeping, and that is a copy and creative fix you can make this week.
Rewrite the two worst product pages. Not the whole catalog. The two items your reason codes just identified. Add the model's measurements or the scale reference, fix the claim that is causing the surprise, and put the fit guidance where somebody will actually read it.
That is a week of work on a number most stores have never examined, sitting directly on top of margin you have already paid for twice.
