Killing the "item not as described" dispute

The "item not as described" (INAD) dispute is the most common and the most preventable returns problem you have, and content verification at receiving is how you kill it. When every returned parcel is photographed and the returned item is checked against what the customer was supposed to send back, swaps, empty boxes, wrong items, and short quantities get flagged at the door, before a refund goes out and before the buyer can escalate to a chargeback. The fix is not better dispute paperwork after the fact. It is capturing decisive evidence at the exact moment the box is opened.

Most INAD disputes fall into a handful of repeatable patterns. Once you can name them and detect them at receiving, an "item not as described" dispute stops being a coin flip and becomes an evidence question you have already answered.

Why "item not as described" is the returns problem worth solving first

An INAD claim is a customer asserting that what arrived, or what they returned, is different from what was agreed. On the outbound side, buyers use it to open a case with a marketplace or card network. On the inbound side, it is the label operators put on a return that does not match the original order. The two are linked: if you cannot prove what came back in the box, you cannot rebut the outbound claim, and you eat the refund plus the return shipping plus the disposition cost. That combined loss is why INAD sits at the center of returns fraud economics, and why it is worth measuring alongside the real cost of a return.

The core failure mode is timing. Traditional returns processing inspects casually, refunds quickly, and only looks closely once a dispute is filed, at which point the item has often already been restocked, resold, or discarded. There is no artifact to point to. Content verification inverts that order: inspect and prove first, refund second.

The common INAD and returns fraud patterns

Nearly every disputed return maps to one of these:

• Product swap. The customer keeps the new item and returns an older, broken, or counterfeit version of the same product, or a different product entirely. This is the classic swap and the hardest to catch without a visual and identity comparison.
• Empty box. The parcel arrives sealed and weighted but the item is missing. Without a photo at open, it is the buyer's word against yours.
• Wrong item. Sometimes fraud, often an honest mistake: a different SKU, size, or color than the order line.
• Partial quantity. A three-pack comes back with two units, or the accessories, chargers, or components are missing. Refund logic that keys off the order, not the contents, pays out in full anyway.
• Wardrobing. The item is worn, used, or installed once and returned as new. It is not missing and it is not swapped, so quantity checks pass. What catches it is condition evidence.

Wardrobing is a grading problem, not a counting problem

Wardrobing slips past swap and quantity checks because the right item in the right count comes back. The tell is condition. When every parcel gets an A/B/C/R condition grade with a confidence score and detected tags at receiving, a used-as-new return grades B or C with visible wear tags, and that grade is timestamped and photo-backed. You are no longer arguing about whether an item was worn. You are showing it.

How content verification flags mismatches at the door

Content verification runs at receiving, before the return is accepted for refund. It answers three questions with evidence attached:

1. Is this the expected item? AI content verification compares the returned item against the expected return for that order and flags swaps and mismatches. A wrong SKU, a different product, or a counterfeit substitution raises a flag instead of clearing.
2. Is the quantity and completeness correct? An automated quantity and completeness check confirms unit counts and that components are present, catching partial returns and empty boxes.
3. What condition is it in? The A/B/C/R grade with confidence captures wardrobing and damage that a match-only check would miss.

Because this happens before the refund decision, a flagged return can be held for human review instead of auto-refunded. That single sequencing change is what converts a future chargeback into a caught mismatch. For the mechanics of putting proof on the record for each parcel, see how we handle photo evidence on every parcel.

Repeat senders surface across returns

One-off fraud is annoying; organized fraud is expensive. Because every return carries an evidence bundle and identity signals, repeat senders surface across returns over time. A buyer who returns swaps or wardrobed items on a pattern becomes visible instead of blending into the noise of thousands of legitimate returns. That lets you set stricter handling, decline pre-refund, or escalate before the next case lands.

INAD dispute types, detection, and evidence captured

How the evidence bundle wins the chargeback

Chargeback defense is a documentation contest. Card networks and marketplaces resolve INAD cases on whichever side supplies clearer, time-stamped proof. Every areturnz return produces an evidence bundle: receiving photos, the AI condition grade with confidence, detected tags, and a full custody chain, available in the dashboard and via signed-JSON API with webhooks. When you respond to a claim, you are not writing a narrative, you are attaching a dated visual record of exactly what arrived and in what state.

Signed JSON matters here. A tamper-evident, cryptographically signed record is harder to dispute than a screenshot, and the custody chain shows the item was tracked from the moment it entered the facility. This is the difference between "we believe the item was worn" and a signed grade of C with wear tags and photos at a known timestamp. To see the artifact itself, look at an evidence sample, and for the operating principle behind it, read why we put proof on every return.

Marketplace context

Marketplaces carry concentrated INAD risk because the platform, not the seller, often adjudicates the case, and buyer-friendly defaults mean weak evidence loses by default. A platform or marketplace that runs returns through a verification node gives every seller the same standard of proof, reduces friction disputes, and can surface repeat-offender buyers across the whole marketplace rather than one storefront at a time. That is the model behind our work with marketplaces.

Putting it in place

Killing INAD disputes is an operations decision, not a policy tweak. The requirements are concrete:

• Photograph every parcel at receiving, without sampling.
• Verify contents against the expected return before the refund fires, not after.
• Grade condition with confidence so wardrobing is caught, not just swaps.
• Store proof as signed, custody-tracked records you can pull for any case.
• Track sender patterns so repeat fraud is visible across returns.

Across 180K+ returns processed at NJ-01 with roughly 99.6% match accuracy and a median cycle around 48 hours, the pattern holds: the disputes you win are the ones you documented at the door. INAD is preventable when verification happens before the refund instead of during the fight.

Frequently asked questions

What is an "item not as described" dispute?

It is a claim that the goods received or returned differ from what was agreed: a swap, a wrong item, a short count, an empty box, or a used item returned as new. It is the most common returns dispute type and the most preventable with content verification and photo evidence captured at receiving.

How does content verification prevent returns fraud before a refund?

At receiving, AI compares the returned item against the expected return and checks quantity and completeness before the refund decision. Swaps, wrong items, empty boxes, and partial returns get flagged and held for review instead of auto-refunded, so the loss never happens.

How do you catch wardrobing if the right item comes back?

Wardrobing passes swap and quantity checks because the correct item and count return, so the tell is condition. An A/B/C/R grade with confidence and wear tags at receiving flags used-as-new returns with photo proof, turning a judgment call into documented evidence.

What evidence actually wins a chargeback?

Time-stamped receiving photos, an AI condition grade with confidence, detected tags, and a full custody chain, delivered as a signed-JSON record. Because it is dated and tamper-evident, it outweighs the unverifiable narrative on the other side of most INAD cases.

How does this help a marketplace specifically?

A marketplace can route returns through a verification node so every seller gets the same standard of proof, disputes are adjudicated on evidence, and repeat-offender buyers surface across the whole platform rather than one storefront at a time.

Related reading: Returns Evidence and Dispute Proof: The Complete Guide