Fraudulent Returns: How to Spot Them With Your Own Data

Almost every seller with volume has lived this scene at least once: the return box arrives, you open it expecting your product, and something else is inside. A generic charger instead of the original. A different sneaker from the one you shipped. A box with the right weight, stuffed with paper. The refund is already processed, the commission is already paid, the outbound and return freight are already charged, and you are left with no product and no money.

The natural reaction is to get angry and generalize: “people abuse this.” That is where the mistake starts, because the generalization pushes you to fight with legitimate buyers, to harden policies you don’t control, and to burn time on arguments you won’t win. The vast majority of returns are not fraud: they are the wrong size, an expectation your listing set badly, ordinary buyer’s remorse, and product that arrived damaged in transit. Treating all of them as suspicious is the fastest way to destroy your standing on the channel.

Abuse does exist, but it is a minority, and the only way to attack it without damaging the rest of your operation is to identify it with data, not with gut feeling. That means knowing which patterns exist, what traces they leave in your own history, how to document them at the right moment, and which official channel handles the claim on each platform.

This article is about exactly that: defending yourself inside the rules, with evidence, without inventing your own policies that the marketplace will not back up.

first, get the proportion right

Before talking about patterns it’s worth sizing the problem honestly, because perception almost always inflates it. The public measurements available are from US retail, not from Mexican marketplaces, and they should be read as an order of magnitude, never as a diagnosis of your business.

With that caveat: the National Retail Federation estimates that roughly 9% of returns in the United States are fraudulent in the strict sense, while broader measurements — which include policy abuse, not only outright fraud — put the figure closer to 15% of returned volume. The same sources report that merchants who keep records saw empty-box and decoy-item returns grow.

As far as I could verify, there is no equivalent public measurement for Amazon Mexico or MercadoLibre Mexico. So the number that matters is yours, and it comes from counting your own confirmed cases — not the suspected ones — month over month.

The practical conclusion of that proportion is uncomfortable but useful: if nine out of ten returns are legitimate, any policy that punishes everyone to catch one is extremely expensive. The real work is not in distrusting more, but in detecting better.

the four known patterns

Return abuse is not infinite. It repeats in four shapes, and each leaves a different trail.

The empty or padded box. The buyer returns the packaging with equivalent weight — books, paper, a piece of scrap metal — so the parcel doesn’t look suspicious at the drop-off point. It’s the most brazen pattern and also the easiest to document, because the weight recorded on the return label almost never matches the real product’s weight.

The swapped product. Something of the same size and weight comes back, but it isn’t your unit: a generic version, a used piece, another model in the same family. The key here is the serial number, the batch, or any mark that identifies the exact unit you shipped.

Use and return. The buyer uses the product — once, for a weekend, for a season — and returns it within the window as if it had never been opened. In apparel it’s known as wardrobing; in electronics and tools it shows up as “I tried it and it didn’t work for me” after obvious use. It leaves physical signals: wear, smell, tags cut and reattached, missing accessories.

Returning a different item altogether. Less common but real: the buyer sends back a product they never bought from you, sometimes from another store entirely, betting that nobody will inspect closely at receiving. You catch it by cross-checking the order against what actually arrived.

To those four, people usually add a fifth case that is not fraud even though it hurts just as much: bracketing, buying three sizes to keep one. It is a behavior the channels’ policies allow, and it should be treated as a structural cost of the category, not as abuse.

Glossary: real net margin, with everything deducted →

the signals you can actually measure

Intuition is a bad detector because it only remembers the outrageous cases. History is better, because it reveals concentrations a person can’t see.

Concentration in one SKU far above your median. If a product returns at three times your usual rate and the recorded reasons are neither size nor defect, that’s where it pays to look unit by unit. Abuse almost always concentrates in products that are expensive, small, easy to resell, or that have a very similar generic version.

A gap between the stated reason and the physical evidence. A return marked “never opened” that arrives with the packaging broken is a data point, not an accusation. What matters is that it gets recorded at receiving, not three weeks later.

Repetition by buyer. Platforms don’t hand you a full buyer profile, but you do see your own orders. A single user who systematically returns the same type of product is worth documenting — always to report through the official channel, never to block on your own.

Rate difference between channels for the same product. This is the most informative cross-check of all. If the same SKU returns at 6% on one channel and 18% on another, the product didn’t change. What changed is the listing, the channel’s policy, the friction to return, or the kind of traffic you are buying. Before suspecting buyers, rule out that the problem is your own listing.

Anomalous seasonality. A spike in returns concentrated right after a specific date — an event, a long weekend, a use season — points to use-and-return, not to a factory defect.

document before you claim

No claim is won with indignation. It is won with evidence collected before the case exists, because afterwards there is no way to reconstruct it.

The minimum viable setup, and it costs almost nothing to run:

• A photo or video of the package as it leaves, with the label visible, for orders above a value threshold you define.
• Serial number, batch, or a discreet mark recorded against the order number. On products where swapping is common, this is the only thing that distinguishes your unit from a generic one.
• Continuous, uncut video of opening returns, with the label and the parcel’s condition visible at the start. It is the evidence that carries the most weight and the one almost nobody has when they need it.
• Recorded outbound weight versus the return label’s weight. A large gap is an objective, verifiable indication.
• A record of the condition the unit arrived in, with the buyer’s stated reason next to it.

That file — boring and routine — is the difference between a claim that goes through and one that collapses. It also has a side benefit: it forces you to separate returns by real cause, which is exactly the data you need to bring the legitimate rate down.

each platform’s official channels

This is where precision matters, because policies change often and vary by category, by fulfillment program, and by country. Everything below must be confirmed in Seller Central and in MercadoLibre’s seller help before you act, because what was true last quarter may not be true today.

On Amazon, the mechanism designed for this is the SAFE-T claim (Seller Assurance for E-commerce Transactions), meant to cover the seller when the buyer returned something materially different from what was shipped, when the unit came back damaged by the buyer’s use, or when a refund was applied that the seller considers incorrect. It is opened from the orders section in Seller Central. Public documentation indicates there is a limited window of days to file it from the moment the refund was applied or the unit was received, and a short deadline to answer the investigator’s information requests. The exact deadlines and the program’s availability per marketplace need to be verified in your own account: they are not identical across countries or across fulfillment models.

On MercadoLibre, the flow runs through the claim inside the order itself and, if it isn’t resolved there, through mediation. Official help states that the platform reviews cases and covers the seller when there is sufficient evidence that the product shows signs of misuse or that the return was made with fraudulent intent, and that a quality check is performed when the product comes back. Same as with Amazon: the criteria, the deadlines, and which categories apply should be read in Meli’s help the moment you need them.

In both cases the logic is identical and worth internalizing: the marketplace decides with whatever you put in front of it. A claim with unboxing video, serial number, and a weight comparison is a different conversation from one that says “they sent me back a rock.”

what not to do

This deserves saying out loud, because frustration pushes people toward shortcuts that turn out expensive.

This is not about denying legitimate returns: beyond being unfair to the buyer, it is the direct route to degraded account health metrics and, if repeated, to suspension. It is also not about inventing your own policies that contradict the channel’s — shorter windows, restrictions the marketplace doesn’t recognize — because they are not enforceable and they do cost you claims. Nor about accusing the buyer in the channel’s messaging: it resolves nothing and it is on the record.

Effective defense is the boring kind: document well, claim through the formal channel, accept what doesn’t qualify, and use the aggregate data to fix what is actually in your hands.

Glossary: inventory valuation, how much capital is sitting still →

how this reads in iqseller

In the Profitability module, a return does not live as a separate report: it enters as a provision inside each SKU’s contribution margin. That product’s historical rate, multiplied by what a full return costs — refund, channel commission already paid, FBA or Full fees, outbound freight, return freight, and the unit that sometimes comes back unsellable — is deducted before deciding whether the product makes money.

That framing changes the question. It stops being “am I being robbed?” and becomes “can this SKU sustain the return rate it has?”. A product with 30% gross margin and an expensive return can be losing money even when none of those returns is fraud.

And the comparison that best separates fraud from a self-inflicted problem — the same SKU’s rate on Amazon versus MercadoLibre — comes out directly, because orders and settlements from both channels land in the same catalog and the same Parent → Model → SKU tree. If the number is even across channels, the product is the problem. If it spikes on only one, the problem is in that listing or that channel, and there is nothing to claim: there is something to fix.

The Alerts module handles the timing side: a SKU that steps outside its own return range flags while there is still time to document, not at month-end close.

the balance you have to hold

Defending against abuse and being easy to buy from are not opposing goals, but they do have to be managed together. Every gram of friction you add to catch the 9% costs you conversion with the other 91%, and that arithmetic almost always comes out negative.

So the sensible sequence is: measure first, then document the full flow with no exceptions, then concentrate your claim effort on the SKUs and amounts where it is worth it, and finally — this is what recovers the most margin — use the aggregate to fix the avoidable returns, which are far more numerous than the fraudulent ones.

Fraud gets claimed. Everything else gets corrected. Confusing the two is what lets returns eat your year.