A return rate you know is too high
One of your products gets returned far more than the rest. The FBA customer returns report tells you the reasons Amazon's codes allow: defective, not as described, no longer needed, wrong size, did not like. Many returns come with a short comment from the buyer. 'Handle came loose after a week.' 'Smaller than the photos suggest.' 'Colour is more grey than blue.' Those comments are the useful part, and they are sitting unread in a spreadsheet column.
Meanwhile the Voice of the Customer page in Seller Central may be showing a warning on the same product, and the reviews are saying something similar in different words. Nobody has put the three sources together.
Why returns are hard to learn from
Return data is plentiful but scattered and messy. Reason codes are chosen by buyers from a short list, often carelessly. Comments are short and inconsistent. Reviews and buyer messages about the same issue use different wording. And each source lives in a different report.
- Buyers pick reason codes that do not match their comment.
- The same problem is described in many different ways.
- Return comments, reviews and messages are never read together.
- Returns are counted monthly, so a rising problem shows late.
- Nobody owns returns analysis, so it falls between product and customer service.
What unexplained returns cost
Each return costs you the sale, the fees Amazon does not refund, the outbound and return shipping and often the unit itself if it cannot be resold. A high return rate can also trigger Amazon's attention through its product quality measures, and a listing flagged for returns may be restricted.
The larger cost is the fix not made. If returns come from a size chart that is wrong, a photo that is misleading or a component that fails, the problem continues on every order until someone identifies it. The information to identify it was there all along.
The returns analysis we build
- A monthly, or weekly, pull of FBA customer returns, FBM returns, reviews and buyer messages for each product through the Selling Partner API and your messaging data.
- Comments grouped into causes by a language model you approve, such as size or fit, colour or appearance, component failure, missing parts, damaged in transit, or buyer changed mind, whatever the reason code said.
- A view per product showing the causes, their share of returns, example comments and how each cause is trending over time.
- Links between causes and the evidence elsewhere: matching reviews, buyer messages and any Voice of the Customer notice.
- An alert when a product's returns rise or a new cause appears, sent to whoever owns the product.
- A supplier summary for defect-related causes, with the comments and quantities, ready to share when you speak to the factory.
| Cause found | Where the fix usually lies |
|---|---|
| Smaller or larger than expected | Listing dimensions, size chart, photos with scale |
| Colour not as shown | Photos and colour naming |
| Part broke quickly | Supplier quality conversation |
| Missing parts | Packing checks at factory or prep |
| Damaged in transit | Packaging |
The grouping is checked by a person before it is used for decisions. A language model is good at reading many short comments, but its groups need a sanity check, especially when a product is new.
What your product reviews look like afterwards
Each month, whoever owns the range gets a short list of products with the main causes of returns and a handful of real comments for each. A product returned for being smaller than expected gets its listing checked and a scale photo added. A product with a failing handle goes to the supplier conversation with evidence attached.
After a change, the analysis shows whether the cause fades. That closes the loop that usually stays open: you make a change, and you can see whether it worked.
The same data helps with new products. If a category of products you sell is often returned for sizing, new listings in that category get extra attention to dimensions from day one.
Could returns be telling you something?
- Return comments are never read systematically.
- One or two products have much higher returns than the rest.
- Reason codes say 'defective' but nobody knows what failed.
- Supplier conversations about quality lack specific evidence.
- You have had a Voice of the Customer warning and were not sure why.