What it means
Customers often choose from a small menu of return reasons, so a buyer may select wrong item to start a request even when the issue is a size description, or may not know whether a device has a defect. Coding accuracy checks what the business can verify without accusing the customer.
Define categories and the evidence hierarchy before sampling, keep the customer stated reason, inspection finding and final operational code as separate fields when they differ, and train reviewers to record unknown rather than force a convenient category. This measure is not the distribution of return reasons or the overall product return rate.
A correctly coded defect does not establish who caused it, whether a refund is owed or whether the manufacturer is liable, and those decisions follow separate policy and legal review. Do not use code quality to restrict a valid return, since eligibility, product condition and consumer rights are separate decisions, and the exercise should support better products and service, not force the buyer to diagnose a fault.
Sample eligible closed returns by product and channel, apply consistent rules and record disagreements, reporting unknown and not-yet-inspected items separately. Store returns may have different evidence from mailed returns, and expensive items may receive more inspection, so a single headline accuracy number can hide one unreliable source of data.
Use a second reviewer for a sample of disputed codes, because if reviewers disagree often the rule needs clarifying before the first coder is blamed. Keep the customer's words, since a dropdown category is a convenience for routing while free text can explain the real problem.
If an item is later inspected, retain both the original selection and the new finding instead of replacing one with the other, because a return reason can change after inspection without proving the customer lied. State whether the audit checks a customer-intent code, a product-defect code or a final operational category, since the same return can involve all three: a coat may be returned because it did not fit while inspection also finds a broken zipper.
Build a small taxonomy that staff can apply consistently, with examples and priority rules for overlapping labels such as damaged, defective and not as expected. Reviewers should be able to mark uncertainty rather than guessing to satisfy a required field, and staff should not remove a valid complaint merely because a defect cannot be reproduced.
Use the proper denominator: if five of a hundred returns cannot be assessed, show 85 supported codes among 95 assessable records and five unknowns, because reporting a flat 85% would confuse coding quality with evidence availability. This measure can also find problems in the return form itself, since many customers choosing other suggests the available choices do not match actual experiences.
Revise the form while preserving comparability notes for earlier periods, keep a record of the version used when past cases are reclassified, and check whether changes in the coding rule explain a trend before claiming product quality improved. When the same product has repeated not-as-described codes, read the underlying customer statements before rewriting marketing copy, because the source may be size, colour, compatibility or delivery damage; compare stable cohorts, preserve the old labels for historical interpretation and aim for a testable product or information fix, not a broad accusation about customers.
In practice
Real-world examples.
Example
A clothing retailer reviews closed returns and finds a buyer who chose wrong item because the size chart on the product page was unclear. The reviewer keeps the customer's wording and records a sizing information issue, not a fulfilment error.
Example
An electronics seller receives a return coded as defective, but inspection finds the device works as designed. The record keeps the customer's stated reason, adds the inspection finding as a separate field and flags the code as not supported rather than overwriting it.
Example
A furniture marketplace compares store drop-offs with mailed returns and finds that most mailed returns are not yet inspected. It reports the not-yet-inspected items separately and avoids a single headline accuracy figure that would hide the difference.
Formula
Calculation
Illustrative coding accuracy = Supported reason codes / Assessable returned-item records reviewed x 100. If 85 of 95 assessable codes match the stated rule, accuracy is about 89.5%; report excluded unknown cases as a separate count.Case study
Seen in the real world.
This illustrative and entirely fictional example follows Cedar Wear, an invented online shop. It reviews 100 closed clothing returns. Eighty-five have codes supported by customer notes and inspection; ten have mismatches and five lack enough evidence. Its assessable coding rate is 85/95, about 89.5%, with five unassessable items disclosed separately.
Watch out
Common mistakes.
- Changing a customer reason silently after inspection.
- Treating a provisional code as a verified cause.
- Dropping unknown cases from every report without showing their number.
Questions
People also ask.
Is the customer selected reason always wrong if inspection differs?
No. Preserve both views; they may describe different parts of the experience.
Does accurate coding mean fewer returns?
No. It means reason records better reflect available evidence.
How should uninspected items count?
Keep them provisional or unassessable under a stated rule.
From the founder's library

Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%Related
