Return Reason Codes: A Taxonomy That Actually Tells You Something
A return reason taxonomy works when it captures three things: why the customer says they returned it, what the warehouse found, and what that combination means. Six to eight customer-facing options is the most a list should have, and an other rate above 10 percent means the list has a hole.
The returns report says 40 percent of returns are for other reasons. The product team asks what other means, and nobody knows, because the reason list was written by whoever set up the returns portal three years ago and never revisited. The data is useless, and the list is the reason.
A return reason code is not a form field, it is the only structured data most shops get about why products come back. Done right it tells you which products are misdescribed, which sizes run small, and which suppliers ship defects. Done wrong it tells you other, forty percent of the time.
Why most reason lists fail
The first failure is length. A list with twenty options gets skimmed, and the customer picks the first one that is close enough. Six to eight customer-facing options is the most a list should have, because that is the most a person will actually read before choosing.
The second failure is vagueness. Options like did not like it and changed my mind are not reasons, they are shrugs, and they tell the product team nothing. A useful reason is specific enough to act on: wrong size, not as described, arrived damaged, does not fit as expected.
The third failure is that the list only captures what the customer says. The customer says wrong size, the warehouse finds the item unworn and correct, and the real story is bracketing, the customer ordered two sizes and kept one. The customer-facing reason is one layer, and it is not the whole truth.
Three layers, not one list
The taxonomy that works has three layers. The customer reason is what the shopper picks at the return. The warehouse finding is what inspection records when the item arrives: condition, completeness, and whether the stated reason matches what is in the box. The derived flag is what the combination means, like bracketing or wardrobing or a sizing problem.
Amazon's scale shows what the structure looks like when it is taken seriously: 33 reason codes and 6 condition codes, with the condition codes doing the work the reason codes cannot. A small shop does not need 33 reasons, but it does need the separation between what the customer said and what the warehouse found.
The layers have to be captured at different moments by different people. The customer picks the reason at the return request, the warehouse records the condition at inspection, and the derived flag is computed from the pair. A taxonomy that asks the customer to do all three layers is a taxonomy that fails.
The numbers that tell you the list is broken
Review the taxonomy quarterly, and retire any code that falls under 1 percent of returns. A code nobody picks is clutter, and clutter is what pushes customers toward other. The review takes an hour and it is the highest-value hour in the returns calendar.
The number that matters most is the other rate. If other is above 10 percent, the list has a hole, and the missing option is usually obvious from the support tickets. Add the missing option, watch other drop, and repeat the review next quarter.
The second number is the mismatch rate: how often the warehouse finding contradicts the customer reason. A high mismatch rate usually means a product problem, not a customer problem, because the customer is describing the symptom and the warehouse is finding the cause.
Related reading
Two earlier posts on this site cover ground “Return Reason Codes: A Taxonomy That Actually Tells You Something” builds on: Return Rates by Channel: Online Runs Triple the Store, Net Return Rate: The Number That Decides Your Margin, and Return Rates by Category: Nineteen Percent Is the Number, Your Category Is the Question.
Frequently asked questions
How many return reason codes should a shop have?
Six to eight customer-facing options is the most a list should have, because that is the most a person will actually read before choosing. Behind the scenes, the taxonomy needs three layers: the customer reason, the warehouse finding, and a derived flag for what the combination means.
What does a high other rate mean in returns?
It means the list has a hole. If other is above 10 percent of returns, the missing option is usually obvious from the support tickets. Add it, watch other drop, and review the taxonomy quarterly, retiring any code that falls under 1 percent.
What is the difference between a return reason and a warehouse finding?
The reason is what the customer picks when requesting the return, and the finding is what inspection records when the item arrives: condition, completeness, and whether the stated reason matches the box. Amazon runs 33 reason codes and 6 condition codes, with the condition codes doing the work the reasons cannot.
How do I detect bracketing from return data?
Bracketing shows up as a mismatch between the layers: the customer says wrong size, the warehouse finds the item unworn and correct, and the order history shows two sizes ordered at once. The derived flag is computed from the pair, not from either layer alone.