Empty-box and wrong-item claims: the returns fraud you cannot see
Every returns team has a file of claims that cannot be resolved by looking at anything: the box arrived empty, the customer says, or the box contained the wrong item, or the item was damaged in a way no photo can confirm or deny. Your warehouse has the box, the customer has a story, and the two do not agree. These are the claims that cost the most per incident and generate the most internal arguments, because someone has to decide who is telling the truth.
Here is the uncomfortable starting point: most of these claims are honest. Warehouses do mis-pack. Carriers do damage boxes. Items do get swapped on shelves. The honest majority is large, which is why a policy of automatic denial would be a customer-service disaster. But a policy of automatic approval is a fraud invitation with a neon sign. Organized abusers know exactly which claim types are unverifiable, and they concentrate there.
The way out is to stop treating each claim as a truth problem and start treating it as a pattern problem. A single empty-box claim from a customer with forty clean orders is almost certainly a warehouse error, and the right move is an instant reship with an apology. The same claim from an account with three previous empty-box claims across two addresses is a different case entirely. The claim is identical. The customer history is what makes them different, which is why per-customer scoring is the only approach that scales.
Weight is your best physical evidence, and most brands do not use it. Carrier scan weights and your warehouse's outbound weights are recorded for almost every parcel. An empty box weighs meaningfully less than a packed one, and that weight is logged at multiple scan points. When a customer says the box arrived empty but the carrier scans show full weight at every handoff until delivery, you have evidence. When the scans show a weight drop mid-transit, the customer is probably telling the truth and the carrier is the problem. Pull weight data into your review process for unverifiable claims and half of them resolve themselves.
Photographs help, but only with structure. Asking for "a photo of the damage" produces a blurry close-up that proves nothing. Instead, ask for two specific photos: the shipping label and the open box with the packing material visible. Fraudsters who planned the claim in advance usually cannot produce a coherent pair, because they did not think to stage them. Honest customers produce them in thirty seconds. The request itself is a filter.
For wrong-item claims, check the pick error base rate. Every warehouse mis-picks at some rate, typically well under one percent for good operations. If a customer's wrong-item claims arrive at five times your warehouse error rate, the warehouse is not the explanation. This comparison is simple arithmetic, but almost no brand does it, because the returns team and the warehouse team report to different people and never compare notes. Put them in the same monthly review.
Design the resolution ladder before the claim arrives. Clean history: instant reship or refund, no questions, with a note that the claim was logged. One prior unverifiable claim: reship, but flag the account for review of the next one. Two or more: manual review with weight data and photo request before anything ships. Documented pattern: require the return of the disputed item before any credit, and apply it consistently. The ladder works because it is automatic. Nobody has to decide in the moment whether to believe someone, which is the decision that generates all the inconsistency.
One more thing: track which SKUs attract these claims. Empty-box and wrong-item claims cluster on high-value, easily resold items. If your $200 jackets generate these claims at ten times the rate of your $40 tees, that is information about where to add friction, like signature confirmation or photo documentation at pack-out for high-risk SKUs. Fraud follows value, and your controls should too.