Return fraud, and the broken orders that look exactly like it

What return fraud is

Return fraud is any use of an e-commerce retailer's returns process to obtain a refund, credit or replacement the shopper isn't entitled to: wearing an item and returning it as new, sending back an empty box, or claiming a delivered parcel never arrived. My read is that it gets counted as a fraud problem and settled as an operations decision. Every behavior on the standard list has an innocent twin, and a brand that can't see what happened to the order is guessing, then charging the guess to whoever it trusts least. The instance I keep meeting is the dull one: an item still in transit long after the rest of the order landed. Left alone it becomes an item-not-received claim someone has to judge blind. We detect it as lost while it's still only late, reship it and file the carrier claim.

The vocabulary isn't settled either: NRF, Target and The Chertoff Group published a retail fraud taxonomy because the industry has no shared language for fraud and abuse.

The behaviors brands actually see

Seven behaviors cover almost everything a brand meets, and longer taxonomies are mostly variants of these.

• Wardrobing: worn or used once, returned as new. The cost lands in resale value, not the refund.
• Bracketing: three sizes ordered, two returned by design. The cost lands in shipping and handling, and it's the policy working as written.
• Empty box and decoy returns: the box comes back empty, or with something else in it. The cost lands in inventory written back in that never existed.
• False delivery claims: scanned as delivered, and the customer says it never arrived. The cost lands in the reship, and this one holds the largest honest population.
• Receipt and price manipulation: a return processed at a price the item was never sold at.
• Stolen goods returns: merchandise never bought, returned for credit. The organized end, and it behaves like crime, not like a customer.
• Serial returning: a shopper whose return rate is an outlier. Not a behavior, a pattern, and the one most often misread.

The split that matters: fraud is deception for gain, abuse is a policy used as written, and only one of the two is the customer's fault. Bracketing and serial returning sit on the abuse side. NRF and Happy Returns found 45% of consumers consider bending return rules acceptable. NRF's theft survey has 55% of retailers reporting rising digital and e-commerce fraud from organized retail crime groups, and 67% recording the involvement of a transnational organized theft group in thefts against their organization. None of it is visible at the RMA step, where the return is received.

The number everyone is quoting is out of date

Almost every guide to this subject leads on a fraudulent-returns share and a headline dollar figure drawn from the 2023 and 2024 editions of one report, and frames the trend as accelerating. The current edition says something else. NRF and Happy Returns put fraudulent returns at 9% of all returns in 2025, against $849.9 billion of total returns, 15.8% of annual sales, with 19.3% of online sales returned. The 2024 edition put returns at $890 billion and 16.9% of annual sales.

The point isn't that fraud halved in a year. The definitions moved and the survey base moved, so the industry doesn't have a stable measurement of how much of this is fraud, and the association producing the number is the one saying so. An operator writing policy off a figure the industry restated downward within twelve months is writing policy off a vibe. Attach the year and the source to every number you quote, this page included.

Where fraud and a broken order look identical

Three of those behaviors have an operational twin that produces the same evidence, and a brand sees evidence, not intent.

• "It never arrived": either the customer is claiming a parcel they received, or it was mis-scanned, mis-delivered or lost. The FBI's Internet Crime Complaint Center logged $503 million in non-delivery and non-payment losses in 2025, and a delivery exception on the same order is usually the tell.
• "The box was empty" or "the wrong item came back": either the customer shipped back nothing, or the order was short-shipped or mis-picked and the customer returned what they were sent.
• "This customer returns everything": either a serial abuser, or a shopper whose size chart is wrong, whose product page photographs badly, or who was sent the wrong item twice.

The difference is never in the return request. It's in what happened to that order between checkout and the doorstep, and in most brands that history lives in a different system from the return. So the decision gets made blind, and a blind decision has two settings: refund everyone and book the loss, or tighten the policy and take the loss from the honest majority. A helpdesk can answer the customer's email about the missing parcel. It can't tell you whether the parcel went missing. Replying about a problem and resolving it are different jobs, and only the second is operations.

What the law fixes in place before you tighten anything

Every guide on this subject recommends tightening the return policy, and none mentions that the policy sits on a statutory floor. New York General Business Law 218-a requires the refund policy to be conspicuously posted and, for online retailers, displayed or hyperlinked before billing information is requested. Where none is posted, the consumer gets 30 days for a full refund or credit, at the consumer's option. California Civil Code 1723 works similarly: anything short of a full refund or exchange for seven days must be posted conspicuously, or the retailer is liable for the purchase amount on a return attempted within 30 days.

Refusing a return doesn't end the matter either. Under Regulation Z, property not accepted by the consumer or not delivered as agreed is a billing error the cardholder may dispute within 60 days of the first statement showing it, and the creditor must resolve it within two billing cycles and no later than 90 days. That clock is what the advice skips: a refused return often becomes a chargeback, so tightening the policy moves the cost rather than removing it.

Measuring it without accusing your best customers

Three measurements turn the fraud call from a guess into a decision:

• Return rate by reason code: read by reason, not in aggregate, because a rising never-arrived rate and a rising delivery-exception rate moving together is an operations problem wearing a fraud costume.
• Refunds issued before inspection: the share settled before anyone opened the box, which is where the empty-box cost lands.
• Time from claim to decision: a disputed return routed to a person is priced at that person's time, which the Bureau of Labor Statistics puts at an annual mean wage of $46,590 for a customer service representative.

The better answer is selective friction: easy for trustworthy shoppers, harder on risky transactions, rather than a blanket tightening. NRF's 2025 returns research has 71% of consumers saying a poor returns experience makes them less likely to buy from that retailer again, so this is a customer retention decision as much as a loss-prevention one.

Deciding the return with the order's history attached

Detect, decide, act. Detect the break when it happens, from signals the systems already emit, not from a return request six weeks later. Decide against the promise that was made and against what already went wrong with this order: late, split, short-shipped, mis-scanned or lost. Act: reship, refund, approve or route to a person, with the reason recorded. One in five orders hits an operational break after checkout, and a break fixed at that point never becomes a return claim at all, because the customer had no reason to make one.

Keeyu doesn't make that call. We don't score returns for fraud, don't represent a merchant to a card network and don't tell you whether a customer is lying: Keeyu isn't a fraud vendor, not a chargeback service, not a returns portal and not a helpdesk. What we act on is the post-purchase operational break, which is what gives whoever does make the call something true to work from.

A return decision made without the order's own history is a coin toss you pay for twice: once in the refunds you should have refused, and once in the honest customers a tightened policy is aimed at. Keeyu is proactive e-commerce operations. We detect the break in the order, decide the fix and act, usually before the customer knows anything went wrong, so the claim either never happens or arrives with its history attached. Every order is a promise. Keeyu keeps the promise.

• National Retail Federation and Happy Returns. 2025 Retail Returns Landscape. Fraudulent returns at 9% of all returns in 2025, 19.3% of online sales returned, and 45% of consumers saying bending return rules is acceptable.
• National Retail Federation. Returns projected at $849.9 billion in 2025. The headline total, 15.8% of annual sales, and 71% of consumers saying they are less likely to shop with a retailer again after a poor experience.
• National Retail Federation and Happy Returns. 2024 retail returns to total $890 billion. The prior year's totals, 16.9% of annual sales, and the basis for the year-on-year comparison.
• National Retail Federation. Retail Fraud Taxonomy. Published with Target and The Chertoff Group because the industry has no shared language for fraud and abuse.
• National Retail Federation. The Impact of Retail Theft and Violence 2025. 55% of retailers reporting increases in digital and e-commerce fraud by organized retail crime groups, and 67% recording the involvement of a transnational organized theft group.
• FBI Internet Crime Complaint Center. 2025 Internet Crime Report. $503 million in reported non-payment and non-delivery losses in 2025.
• New York State Senate. General Business Law 218-a. The posted refund policy, the online display or hyperlink before billing information is requested, and the 30 day default where none is posted.
• California Legislative Information. Civil Code 1723. Conspicuous posting where the policy is anything short of a full refund or exchange for seven days.
• Consumer Financial Protection Bureau. Regulation Z, 12 CFR 1026.13. Property not accepted or not delivered as agreed as a billing error, the 60 day notice window and the two billing cycle resolution limit.
• US Bureau of Labor Statistics. Occupational Employment and Wage Statistics, May 2025. Annual mean wage of $46,590 for customer service representatives.