Answer in brief
Every return arrives with a reason, and the reason points somewhere specific: the listing copy, the product itself, or the box it travelled in. How to tell them apart, what to change first, and why a lower return rate is usually a copy and photography job.
The short answer: where to start when returns pick up
The first useful move with a return is to read it rather than write it off. Every return carries a reason, and that reason points at one of three places: the listing set the wrong expectation, the product did not survive use, or the packaging did not survive the trip. Until you know which, every fix is a guess.
The second move is to start with the listing, not the factory. Changing a supplier is slow and expensive, while reshooting one frame and rewriting three lines of description takes days. Whatever share of your returns comes from expectation is the share that clears fastest and costs least to clear.
The third is to count returns by reason instead of watching one percentage. A single rate tells you something is wrong without telling you what. A breakdown by reason turns a loss into a work order: one line goes to the copy, one to the camera, one to the packing bench.
The rest of this piece is about separating those three sources, deciding what to change first, and knowing whether the change worked. Sequence matters more than effort here, because a misread cause makes you pay twice for the same problem.
Three addresses: the listing, the product, the box
A listing-caused return looks like this. The item works, it matches the description as the seller reads it, and the buyer still did not want it. The stated reason sounds like not what I expected, smaller than the photo, or a different shade from the one on screen.
A product-caused return reads differently. The item arrived intact and then failed: a seam opened, a coating wore off, a fabric stretched, a function stopped. No amount of rewriting helps here, because the fault sits in the material, in the construction, or in one particular batch.
A packaging-caused return is the easiest of the three to spot. Damage is visible from the outside: a crushed corner, a leak, a dented box. The item was fine in the warehouse and stopped being fine in transit. That is an engineering problem, not an editorial one.
The boundaries blur, and some returns honestly belong to two addresses at once. Even a rough split beats a single counter, because each group is fixed by different people, on a different budget, over a completely different timeline.
Reading the reason when the buyer wrote nothing
Most buyers pick a reason from a short and generic dropdown. Wrong size can mean the garment genuinely runs small, or it can mean your measurement table was buried in the sixth gallery image, where nobody looked before placing the order.
So you read reasons by repetition rather than one at a time. Three returns with identical wording on the same size are a signal about a specific line in your listing. One return, on its own, is a person having an ordinary day.
Read returns alongside buyer questions and the text of reviews, including reviews from people who kept the item. A question asked repeatedly marks the place where the listing goes quiet, and listing silence has a habit of ending in a return a fortnight later.
Your own pre-sale messages are a third source. When someone on your side explains the same detail to every buyer who asks, that explanation belongs in the description, not in a chat window where exactly one person ever sees it.
Expectation returns: when the listing is at fault
Expectation is not built by one element. The title, the first frame, the specification table and the price get assembled by the buyer into a mental picture, and when the parcel fails to match that picture, a perfectly functional item comes back at your expense.
The first mechanism is omission. Sellers leave out whatever feels obvious to them: the stiffness of a sole, the weight, the smell of a new material, the fact that assembly is required. What is obvious to the manufacturer is rarely obvious to someone holding the thing for the first time.
The second mechanism is flattering imprecision. Rounded dimensions, optimistic colour names, vague wording about what is actually in the box. That imprecision wins one sale and buys a return that costs more than the sale you were trying to protect.
There is a cheap test. Read your own listing as someone who knows nothing about the product, and write down every question it leaves unanswered. Each unanswered question is a candidate reason for a future return, and most of them close with a single sentence.
Size, fit, and everything a screen will not let you touch
Categories that involve trying things on return more by design. People order two sizes in order to keep one, and that is the normal shape of the purchase rather than a broken process. Policing it does not work; removing the need for it does.
Precision is the lever. Measurements of the actual item rather than a brand-wide chart, the height of the model and the size worn, a comparison with an object the buyer already owns: each fact removes one reason to order a spare size on speculation.
Naming the quirks helps as well. Runs small, stretches with wear, shrinks after washing, narrow last. Sellers avoid those sentences because they sound like warnings, but they filter out the wrong buyer before shipping instead of after delivery, which is the cheaper end of the process.
Keep measurements and attributes as text rather than as artwork. Text is indexed, read aloud by assistive software and carried into other languages. A number baked into a JPEG does none of that, and it disappears entirely the moment the image fails to load.
Photography as a promise: where the shoot creates the return
Photographs promise more than text because they are read faster. A colour shot under warm light is taken as the colour of the item; a texture pushed in post-production is taken as the real texture. The gap between the frame and the parcel gets filed as a return.
The working rule is to shoot so the parcel looks no worse than the frame, never the other way round. One honest frame in neutral light, one on a person or in a hand, and one with a scale reference will answer most of the questions that otherwise turn into orders placed on hope.
A note on fully model-generated images. Beyond the risk of showing details your product does not have, in a number of jurisdictions an image with no human authorship may leave nobody holding exclusive rights, which means a near-identical frame can lawfully appear on a competitor listing.
Photography is not included in any VITON13 marketplace package: not in the Demand Audit, not in Listing Refresh, not in the Marketplace + SEO System. We specify which frames are needed and why, and you or your photographer take them.
When the return points at the product, not the copy
Some returns cannot be written away. The item arrived intact and failed within a week: a button sank, a seam opened, a coating rubbed off. It is cheaper to admit the fault sits in the object than to spend another round of budget on the description.
These returns usually cluster, by batch, by colour, by size, or by delivery date. A match along any one of those axes is a reason to pull one specific batch rather than condemn the whole line and open a sourcing project that nobody actually asked for.
While the cause is unresolved, selling with the limitation stated plainly beats selling in silence. Silence converts a shortcoming into something the buyer experiences as misrepresentation, and that comes back as a review, which is far more expensive than a single return.
Sometimes the right answer is to pause the listing. That looks unpleasant on a spreadsheet and is still cheaper than letting a listing accumulate a history you will then spend months of promotion and discounting trying to outrun.
Packaging, transit, and damage that appears on the road
Transit damage is rarely random. The fragile point tends to repeat: the same corner, the same neck of the bottle, the same seam around the lid. What matters is not the total count of broken parcels but the place where the damage keeps landing.
Marketplace logistics is built for throughput, not for your particular item. Boxes get dropped, stood on end and stacked under heavier ones. Packaging has to survive that scenario, not the careful hand-to-hand delivery you picture while packing the very first sample by hand.
Cheap fixes work more often than sellers expect: a better-fitting box, void fill, a different orientation inside, one reinforced corner. Expensive fixes, such as moving to a whole new format of container, deserve consideration only after the cheap ones have been tried and failed.
Packaging also shapes expectation. A battered box lowers the perceived quality of a working item and raises the chance the buyer sends it back under some other wording. The record will say did not like it, while the actual cause stayed on the road.
What to change first: the order that pays back soonest
The order of work is set by the speed and the cost of testing, not by importance. First come the things you can change in a day and see immediately: the title, the opening lines of the description, the attribute fields, the measurement table. The cost of being wrong here is close to nothing.
Next is the gallery. Which frame leads, whether scale is visible, whether the buyer can see everything that is actually in the box. This costs more than text and remains incomparably cheaper than restarting production or changing suppliers.
Packaging comes third, because a change requires purchasing and a fresh shipment, but it measures cleanly: same product, same listing, different box. Very few experiments in ecommerce isolate their variable that neatly, so the answer arrives without much argument.
The product changes last, not because it matters least but because the cycle is long. While that cycle runs, the first three steps are already removing the returns that were never the product's fault to begin with.
Checking whether the change actually worked
Judge a change by the group of returns it was supposed to close, not by the headline rate. If you rewrote the sizing block, look at size-related returns, because the noise from every other reason will otherwise bury the result completely.
Changing everything at once is convenient and unreadable. One or two changes per cycle produce an answer you can carry over to other listings. Ten at once produce a new headline number and no explanation whatsoever for why it moved.
Wait a full cycle: delivery, use, and the return itself being filed. Comparing the week before with the week after tells you nothing, because part of the relevant returns has not physically travelled back to your warehouse yet.
Write down what you changed and when. Three months later, without that note, there is no way to separate the effect of an edit from a season, from a discount, or from a shift in how the marketplace ranks the listing in search.
Product data and accessibility belong to the same job
Expectation is not set only on the marketplace. The same product lives on your own site, and there the attributes are worth marking up as data. The Product type in Schema.org carries size, color, material and weight, plus arbitrary attributes through additionalProperty.
The same vocabulary includes hasMerchantReturnPolicy, pointing at a MerchantReturnPolicy with a return category and a window. Markup does not reduce returns by itself; it removes the discrepancies between what your site says and what the storefront shows, and discrepancies are where expectation breaks.
Accessibility pushes in the same direction. WCAG 2.2 asks for text alternatives for non-text content and steers you away from images of text wherever real text can do the job. Measurements baked into a picture do not reach every buyer who needs them.
Contrast is the second practical item: 4.5:1 for ordinary text and 3:1 for large text. Pale grey lettering over a bright photograph is invisible to a share of your buyers, and a caption nobody reads works exactly like a caption you never wrote.
What VITON13 does here, and what it costs
The Demand Audit is $50, 1-2 working days, with one round of revisions on the findings document. It sets out which expectations your listings currently create and where the copy drifts away from the product. Access to your seller account is not included.
Listing Refresh is $70, 2-3 working days, one round of revisions. Titles, descriptions and attribute fields get rewritten against what the product actually is. Photography, ad spend and account management are not included in that package.
The Marketplace + SEO System is $110, 3-5 working days, with two rounds of revisions on listings and copy. Marketplace + SEO Express compresses the same work into $190, 3 working days and one round. Photography and paid placement sit outside both of them.
Managed Demand Growth runs at $200/mo on a monthly cycle, with 30 days notice to stop, and advertising spend on the marketplace is not included in the fee. The full scope and the briefs are published at /services/marketplaces.
Practical checklist
- Sort the last month of returns into three addresses: listing, product, packaging.
- Write out the repeating reason wordings and find the listing line responsible for each one.
- Replace the brand-wide size chart with measurements of the actual item and add a scale reference to the photos.
- Check that measurements and box contents live in text, not only inside an image.
- Find the repeating point of transit damage and reinforce the packaging exactly there.
- Record the date of every edit and compare it only against its target group of returns.
Questions and answers
Where do I start if my return rate has jumped?
With a breakdown by reason for the last month. While returns sit as one number, you cannot tell whether to fix the copy, the product or the box, so every edit is a guess.
How do I know the listing is at fault rather than the product?
The item works, and it matches the description as the seller reads it, yet the buyer says they expected something else: another size, another shade, another set of contents. That is the listing address.
What does a return-cause analysis cost at VITON13?
The Demand Audit is $50, 1-2 working days, with one round of revisions on the findings document. Access to your seller account is not included in that package.
Is product photography part of your packages?
No. Listing Refresh excludes photography, ad spend and account management, and the Marketplace + SEO System excludes photography and paid placement. We specify the frames; you shoot them.
How long before I can judge the result of an edit?
Not before a full cycle has run: delivery, use, and the return being filed. Comparing the week before with the week after almost always shows noise rather than effect.