A return is a bug report. You pay the postage.

Field notes on reading the returns column as catalog telemetry, learned later than we'd like to admit.

The most detailed product feedback your store will ever receive arrives in a cardboard box. You pay the shipping on it, and most operators never read past the refund amount.

We treated returns like weather for years. Seasonal, unpleasant, nobody's fault (we audit catalogs for a living and still didn't think to open the returns tab).

The weather theory

The weather theory is comfortable because every part of it blames someone else. Carriers crush boxes. Serial returners game policies. Shoppers bracket sizes because the sofa is a fitting room now.

Some of that is real, and the scale is real. NRF and Happy Returns expected American retailers to take back $849.9 billion in merchandise in 2025. Online, 19.3 percent of sales make the round trip, roughly one order in five.

Fraud exists too: NRF pegs 9 percent of returns as fraudulent. Which means the other 91 percent are telling you the truth.

Then we read the reason codes

A return reason is a form customers fill out for free while mildly angry. It is the one dataset in the business where nobody is being polite.

Coresight Research traces 53 percent of apparel returns worldwide to fit, ahead of every other reason code. Invesp puts another 22 percent of online returns on products that looked different from the picture.

Then Akeneo surveyed 1,800 consumers across eight countries, and 40 percent had returned a purchase because the product information was simply wrong: sizing, images, specs.

Read those codes again, slowly. "Doesn't fit" is a size chart that lied or never existed. "Looked different" is a photo shot for another variant, or a description written by someone who never touched the product.

Those aren't logistics failures. They're catalog defects, found in production, by paying customers.

Bug reports with tracking numbers

A software team would recognize the format immediately. Reproduction steps: the order. Expected behavior: the product page. Actual behavior: the thing in the box. Your customer wrote the ticket, and you paid return shipping to receive it.

The customer is the most expensive QA department you will ever staff. Most returns programs optimize the apology instead of the cause: faster refunds, nicer portals, softer emails, while the size chart that generated the ticket stays wrong.

Bracketing is the same signal, louder. NRF found 51 percent of Gen Z shoppers buy multiple items intending to return some. That is a customer building a private fitting room at your expense, because the size data on the page gave them nothing to trust.

Nerd detail, skip freely: on the catalogs we open, size information mostly lives in a JPEG, a screenshot of a spreadsheet pasted into the description (occasionally photographed, as far as we can tell, from across the room). A size recommender can't parse it, semantic search can't index it, and an AI shopping agent will walk right past it. Structured fit data means measurements in actual fields, per variant.

What moves when the data moves

When Gunner Kennels added 3D models and AR to its product pages, Shopify's own write-up records the return rate dropping 5 percent. No policy change involved. Customers finally saw the crate they were buying.

That is the whole thesis in one case. Returns are a P&L line, and part of that line is rented by bad product data. Fix the data and the line moves; apologize faster and it doesn't.

We process catalogs for a living, 642,000+ SKUs so far, and the returns column is now among the first things we ask for when a client wonders where the margin went. It is a catalog audit that already ran, at full price, in production. You funded it.

What the boxes taught us

• Pull your return reason codes before you touch your returns policy. The policy is rarely the defect.
• Treat every "doesn't fit" as a size data bug until proven otherwise.
• Fit data belongs in structured fields per variant, not in a JPEG. Assign it like catalog work, because it is catalog work.
• If the same reason repeats on the same product, the page is wrong, not the customer.

The boxes keep arriving either way. Read them.

If you'd rather find these defects before your customers ship them back, OKART's Infrastructure Stress-Test reads your catalog the way the returns column eventually will, read-only, delivered in 48-72 hours.

Sources

1. Consumers Expected to Return Nearly $850 Billion in Merchandise in 2025 (NRF): $849.9 billion projected 2025 returns, 19.3% online return rate, 9% of returns fraudulent.
2. NRF and Happy Returns Report: 2024 Retail Returns to Total $890 Billion (NRF): 2024 baseline figures and 51% of Gen Z consumers bracketing purchases.
3. Industry Benchmarks for Clothing Return Rates by Product Category and Country (PRIME AI): Coresight Research figure of 53% of apparel returns attributed to fit issues.
4. The Most Common Ecommerce Return Reasons (Corso): Invesp figure of 22% of online returns caused by products looking different than expected online.
5. 40% of Consumers Returned Products Due to Incorrect Information (360 Magazine, on Akeneo research): Akeneo survey of 1,800 consumers in eight countries; 40% returned due to inaccurate product information.
6. Ecommerce Returns Management: How To Reduce Returns (Shopify): return-rate context and the Gunner Kennels 3D/AR case with a 5% return-rate reduction.