Return Reason Codes: Building a Taxonomy Your Data Can Actually Use
Direct answer: A usable taxonomy has a short list of top level reasons a customer or agent can pick in seconds, each with a small set of sub reasons that name the fixable cause. Changed mind, wrong size, damaged, not as described, and late delivery cover most volume; sub codes separate product problems from description problems from logistics problems, which is where the actionable data lives.
Every returns operation sits on reason code data, and most of it is close to useless. Too much volume lands in other, size complaints mix product sizing errors with customer ordering errors, and damaged covers everything from crushed boxes to dead electronics.
This guide rebuilds the taxonomy from the decisions it has to feed: what product teams fix, what content teams rewrite, and what logistics changes. If a code cannot change a decision, it should not exist.
Start from decisions, not from labels
List the decisions first. Product teams need to know which items run small. Content teams need to know which descriptions mislead. Logistics needs damage by carrier and packaging type. Design codes that answer those questions and stop there.
The classic failure is designing codes around how returns are processed instead of why they happened. Warehouse handling categories and customer cause categories are different taxonomies, and forcing one list to serve both corrupts both.
The two level structure
Level one is what the customer sees: a short list, usually six to nine options, in plain language, pickable in seconds. Accuracy at level one matters more than detail because it routes everything downstream.
Level two is assigned by rules or by the receiving team from inspection: under wrong size, for example, runs small, runs large, ordered multiple sizes, and size chart wrong are different fixes. Inspection assigned sub codes are far more reliable than asking customers for finer detail than they can judge.
The codes that earn their place
Not as described must split into description wrong, photos misleading, and missing feature information, because each has a different owner. Damaged must split transit damage from product defect, since one bills the carrier and the other bills the supplier.
Changed mind deserves one honest sub code most taxonomies avoid: bought to try, or bracketing, where customers order multiples intending to return most. Measuring bracketing openly is the first step to policies and tools, like better fit guidance, that reduce it.
Keeping the taxonomy alive
Audit code distribution monthly. If other exceeds about five percent, sample those returns and find the missing code. If a code never changes a decision in a quarter, merge it upward. Taxonomies rot by addition; every new code needs an owner and a decision.
Finally, close the loop visibly. When a size chart fix cuts wrong size returns on an item, tell the teams whose data made it happen. Reason codes survive on demonstrated usefulness, and each visible win raises the care taken at intake.
Related reading
Related reading: Returns Data Cleanup: Fixing the Records Your Decisions Depend On, Gift Return Workflows: Handling Returns From Someone Who Did Not Buy, and Who Pays for Return Shipping: Allocating the Cost Without Killing Conversion.
Frequently asked questions
How many return reason codes should a store have?
About six to nine customer facing codes, each with two to five sub codes assigned by rules or inspection. Fewer loses the actionable distinctions; more destroys selection accuracy, because people pick the first plausible option when lists get long.
Why is other such a problem in returns data?
A large other bucket means the taxonomy missed real causes, and those returns carry no signal. Sampling other returns monthly and promoting recurring causes into real codes is the standard fix. Aim to keep other under five percent of volume.
Should customers pick sub codes too?
Generally no. Customers can reliably report the headline reason but not the root cause distinctions a business needs, like whether a size problem came from the chart or the garment. Inspection and rules assign sub codes more accurately and keep the customer flow short.
What is the most underused returns metric?
Reason by item over time, watched for sudden shifts. A spike in not as described on one product usually means a listing changed, a supplier changed, or a batch is wrong. It is an early warning system most stores already collect and rarely graph.