How to reduce customer returns, and the half nobody works

What reducing customer returns actually means

Reducing customer returns means cutting the number of orders that come back by removing the causes that produce them, not by making it harder to send something back. Two families of cause do the work: what the customer expected and didn't get, and what your operation shipped and shouldn't have. Almost every list of ways to reduce returns is a list of ways to improve a product page. That work is real, and it has no effect on the returns your own operation created, which is the half that was already preventable and already visible to you. It's the ordinary case, not the exotic one: a customer orders two items, one is picked wrong, the box arrives, and the whole order goes back. No photograph would have stopped that, and the floor knew before the customer did.

The scale isn't in dispute. NRF and Happy Returns put 19.3% of online sales returned in 2025.

Why better product pages only reach half of it

Ask which lever can prevent a given return rather than who was at fault. For one population the lever sits on the product page, before money changes hands. For the other it sits in the warehouse and the carrier network, where no amount of merchandising reaches.

• Expectation returns: the product arrived as ordered and wasn't what the customer wanted: fit, color, scale, feel, a description that promised more than the item delivers. Prevented before checkout, by information.
• Execution returns: the order itself was wrong: wrong item, wrong quantity, damaged in transit, a substitution nobody mentioned, a delivery that missed the date it was bought for. Prevented after checkout, by operations, and never by a better photograph.

Most published advice concentrates its tactics on the first family and gives the second almost nothing. Follow that list and you're optimizing one half of your return rate and leaving the other half alone. One in five orders hits an operational break after checkout, and every one of those breaks is visible in your own systems before the customer feels it: it can be detected, decided on and acted on while the order is still moving. That's the half a customer experience optimization program that stops at the product page never reaches.

The pre-purchase work, done honestly

Do this work. It prevents real returns, and skipping past it would be a dodge. Here's the consensus list with what the consensus leaves out: what each item can and can't reach.

• Description accuracy: say the material, the dimensions and what is in the box. For textiles and apparel that's not presentation, it's an obligation: fiber content, country of origin and manufacturer identity are required disclosures.
• Images at true scale: the single most repeated tactic in the category. It reaches perceived size and color, and nothing else.
• Size and fit data: the one expectation return with a genuinely technical fix. Published work on flagging size bias across a catalog shows the population can be modeled and reduced at scale.
• Real reviews: they only work if they're real, and that's now a rule with teeth: 16 CFR Part 465 prohibits fake reviews, undisclosed insider reviews and suppressed ones.
• A written return reason on every request: coded to a cause rather than a feeling. It's the only input the rest of this article has.

Then the honest limit: every item above changes what the customer knows before they pay. None of them changes what leaves your warehouse.

The returns you can still stop after checkout

From the pick list printing to the customer opening the box, the outcome isn't fixed. That window is where an execution return gets built, and where one can still be stopped.

The prize for stopping them is bigger than the freight saved. When 'where is my order' issues are prevented, shoppers get what they bought on time, as promised, and a return that was really a delivery failure never happens.

• The wrong item picked: caught at pack, or on the weight discrepancy your own systems already record, it's a corrected shipment. Caught by the customer it's a return, a refund and a replacement order.
• A short or split shipment nobody announced: the customer counts the box against the order confirmation. Tell them first and the second parcel is expected rather than missing.
• A stalled or exception-flagged parcel: the carrier posts a delivery exception days before the customer thinks to ask.
• A delivery that will miss the date it was bought for: an order bought for an event is a return the moment it lands late, which is why on-time delivery belongs in a returns conversation.

That last one is a legal obligation as well. Under the FTC's prompt delivery rules a seller must ship in the time it promised, or within 30 days where it promised none, and must offer the buyer a choice between the delay and a refund. The operational answer is the same as the legal one. Nor is it guesswork: return probability can be estimated before the order is even placed.

Seeing that window at all takes deliberate visibility. We create it on three things: a first-in-first-out tracker for orders, a delivery tracker, and a returns tracker. The returns tracker is what makes a preventable return visible while it is still preventable.

There is a number attached to doing this properly. EHP Labs cut tickets by 55%, close to half a million dollars over twelve months against roughly fifty thousand paid to us, which I gave on Marketing for SMEs. Those tickets were operational failures caught before the shopper felt them, and a share of them would otherwise have arrived back in a box.

Reduced is not the same as suppressed

Much of what gets filed under reducing returns reduces nothing. A shorter return window, a restocking fee, exchange only, a return route the customer has to hunt for: each moves the measured number, and none of them stops a wrong item leaving the warehouse. Three things follow.

The customer notices. NRF finds 71% of consumers less likely to shop with a retailer again after a poor returns experience, and the same research has 82% calling free returns an important consideration. A rate bought with friction is paid for in repeat purchase.

There's a ceiling on how restrictive a policy can quietly become. The posted policy governs, which cuts both ways: in California a seller whose policy is more restrictive than a full refund within seven days must post it conspicuously or is liable to a buyer returning within 30 days.

And a suppressed return is still a return in the world. The item was made, shipped and often discarded anyway. Textiles and durable goods already enter US municipal solid waste at scale.

Measure the cause, not the rate

A return rate is one number averaging two different problems together, so it can't tell you which of them you fixed. Segment it, and two segments carry most of the answer.

• Returns with an upstream operational event: returns on orders that carried a pick correction, a damage report, a carrier exception or a late delivery. This is the number a product page can't move.
• Returns per SKU against orders of that SKU: the segment that tells you whether the problem is the item or the operation. One SKU well above the catalog average is a product question. A whole shift or site above it's not.

Give each an owner and a threshold that triggers an action, not dashboard decoration. A rising count of RMA requests on orders that already had an exception attached is a fulfillment defect, not a returns problem.

Cause has a count you can watch. We monitor around eighty points of failure across the post-purchase journey, a number I gave on eCommerce Australia, and the ones that produce returns sit upstream of the return: a mispicked line, a delayed dispatch, an item that was never really in stock. A rate tells you how much came back. Those tell you why.

Preventing returns is an operations job

Most brands route all of this to a helpdesk, which can't do it. A helpdesk answers the customer after the return request exists. It doesn't stop the wrong item leaving the building, re-ship the missing half of an order or warn a buyer that a parcel will be late. It's a system for replying about problems, not for resolving them.

The work belongs to proactive e-commerce operations: detect, decide, act. Detect the break at the pick, the pack or the carrier scan, decide the resolution against the policy, and act before the customer has a reason to send anything back.

Be clear about the edges, because half the answer here isn't ours. We don't write product descriptions, shoot photography, build size charts, render 3D assets or run a review program. Keeyu isn't a returns portal, not a carrier, not a WMS and not an OMS. We don't pick or pack your orders. We read the signals your warehouse and carrier systems already emit, and act on them. The other half is ours: every order is a promise, and an execution return is a promise broken before the box came back.

Stop the returns you are creating

The wrong pick, the split shipment nobody announced, the parcel stalled at a depot, the delivery that missed the date it was bought for: every one of those was visible in your own systems before the customer asked for a refund. Keeyu watches orders against what they were promised, detects the break, decides the resolution and acts on it, usually before anyone has a reason to send a box back. Book a demo and bring your return reasons with you.

The clearest example I have is one I caused. A high-velocity sale day broke the sync between storefront and warehouse, a thousand shoppers paid for stock that did not exist, and every one of those became a refund, an apology and a customer with a reason to shop elsewhere, which I told on The Breakout CEO. Nothing about the product caused that.

Related reading

• Instant refunds, and the return that stays open
• Reasons for returning an item, and which ones are yours
• Ecommerce returns: the half of your return rate you caused
• Post-purchase operations, explained

• National Retail Federation and Happy Returns. 2025 Retail Returns Landscape. An estimated 19.3% of online sales returned in 2025, and 82% of consumers calling free returns an important consideration.
• National Retail Federation. $850 billion returned in 2025. Consumers were expected to return nearly $850 billion of merchandise, and 71% of consumers say they are less likely to shop with a retailer again after a poor returns experience.
• National Retail Federation. 2024 retail returns total $890 billion. The prior year in the same research series, for anyone checking the trend behind the 2025 figures.
• National Retail Federation. Customer returns in the retail industry. The research hub the returns landscape reports are published under.
• Federal Trade Commission. Mail, Internet, or Telephone Order Merchandise Rule. The FTC's business guide to it. A seller must ship within the time promised or within 30 days, and must obtain the buyer's consent to a delay or offer a refund.
• Federal Trade Commission. Selling on the internet: prompt delivery rules. What a seller has to do when it cannot meet the delivery date it advertised.
• Electronic Code of Federal Regulations. 16 CFR Part 435. The rule text behind the prompt delivery obligation cited above.
• Electronic Code of Federal Regulations. 16 CFR Part 465, consumer reviews and testimonials. Fake reviews, undisclosed insider reviews and suppressed reviews are prohibited.
• California Attorney General. Refund policies. There is no general legal right to a refund for a change of mind, so the posted policy is what governs.
• California Legislative Information. California Civil Code section 1723. A retailer with a policy more restrictive than a full refund within seven days must post it conspicuously, or is liable to a buyer returning within 30 days.
• Kedia, Madan and Borar. Predicting returns before purchase. Early bird catches the worm: predicting returns even before purchase in fashion e-commerce. Return probability can be estimated before an order is placed, which is what makes preemptive action possible.
• Nestler and others. SizeFlags: reducing size and fit returns. Published on size and fit related returns in fashion e-commerce. Size and fit returns are a distinct population that can be modeled and reduced across a large catalog.
• Federal Trade Commission. Labeling requirements under the Textile and Wool Acts. Fiber content, country of origin and manufacturer identity are required disclosures, so product page accuracy is an obligation and not only a tactic.
• US Environmental Protection Agency. Facts and figures about materials, waste and recycling. Textiles and durable goods entering US municipal solid waste at scale. Cited for the general waste point.
• The Breakout CEO #86, The Pivot This Founder Made After an Investor Called It Impossible - one in five US shoppers not getting orders on time, and the ticket ratios behind it.
• Add To Cart #578, Stop Customer Problems Before They Start: Inside Keeyu's AI Ops - stuck returns, failed payments and orders caught mid-flight.
• Give it a Nudge, Keeyu is Solving the "Where’s My Order?" Nightmare - why half of all support tickets are post-purchase.