# Recommendations after a return: what to suggest when the first choice failed | KilterMuse Source: https://kiltermuse.com/blog/2026-10-24-recommendations-after-a-return/ KilterMuse answers October 24, 2026 **Short answer:** Post return recommendations should start from the return reason. A size return suggests the same product in a corrected size or a better fitting alternative. A quality or expectation return suggests a different product with the gap addressed. A preference return suggests exploring a neighbouring style. Suppress the exact returned item where the reason implies it will fail again, keep a human tone that acknowledges the miss, and measure success by kept purchases, not clicks. ## The reason is the recommendation brief Returns data is unusually honest feedback, because the customer handled the product and acted. Use the coded reason and, where given, the free text, as the first input to the next suggestion. Too small and too large are opposite briefs that a system ignoring returns will treat identically, by recommending more of the same range in the same size that already failed. Where reasons are unreliable, infer carefully from behaviour: an exchange for another size is a clear signal, while a refund with no comment is weak evidence of anything except that this item, for this person, did not work. Do not build confident personalisation on weak signals. Suppressing the returned item and broadening the suggestions is the honest response to uncertainty. ## Timing and tone rebuild or lose trust A recommendation sent while a return is still travelling, pitched as if nothing happened, feels careless. Sequence post return messages around the actual journey: acknowledgement, refund or exchange confirmation, then suggestions that reference the lesson, such as other fits customers chose after returning this style. That reference, when true, is persuasive because it is specific and helpful rather than flattering. Avoid blaming language and avoid excessive apology discounts as the default. Some customers simply chose wrong and want guidance, not compensation. A discount offered for every return trains returns as a route to a better price and clouds the data you are trying to learn from. Save goodwill gestures for failures that were genuinely the stores fault, such as description errors. ## What to suggest, and what to suppress Suppress the exact item and size where the reason was fit, quality, or not as described, unless the customer asks to see it again. Suggest, in order: the corrected version of what they wanted, the strongest alternative addressing the stated problem, and one exploratory option in a neighbouring style for preference returns. Keep the set small. A wall of alternatives after a failure reads as indifference to what went wrong. Use kept rate data to rank. An alternative frequently bought and kept after this return reason is a proven remedy, and it should outrank a generally popular product. Build those remedy paths deliberately for your highest return products: this size runs small, here is the fit customers kept. That is merchandising knowledge turned into a system rule, and it improves with every return processed. ## Measure kept, not clicked Judge post return recommendations by whether the next order is kept, by time to a successful purchase, and by whether the customer buys again at all. Click and conversion rates alone will reward suggesting the same tempting product that gets returned again, optimising the system toward a loop of orders and returns that looks busy and loses money and trust. Review the top return reasons monthly with the recommendation rules in the same meeting. New failure patterns, such as a changed supplier fit or a misleading photo on one product, should change both the product page and the post return suggestions together. The return was the customer telling you something. The recommendation is where they find out whether you listened. Keep reading: [Recommendations for first time visitors: useful suggestions when there is no history to personalise from](https://kiltermuse.com/blog/2026-10-24-recommendations-first-time-visitors-no-history), [Price sensitivity in recommendations: when showing the expensive option helps and when it kills the click](https://kiltermuse.com/blog/2026-10-24-price-sensitivity-recommendations), and [Replenishment recommendations: reminding repeat buyers at the moment they actually run out](https://kiltermuse.com/blog/2026-10-22-replenishment-recommendations-repeat-buyers) cover the nearby cases in this series. ## Frequently asked questions ### What should a recommendation system do with return reasons? Use them as the brief for the next suggestion. A size return points to a corrected size or better fitting alternative, a quality return to a different product, and a preference return to a neighbouring style. ### Should the returned product ever be recommended again? Only in a corrected form the customer wants, such as another size on an exchange path. Where the reason was fit, quality, or description, suppress the exact item and size by default. ### When should post return suggestions be sent? After the return journey is acknowledged and the refund or exchange is confirmed, not while the parcel is still in transit and the customer has heard nothing. Sequence around the real journey. ### How should success be measured for post return recommendations? By kept purchases, time to a successful kept order, and repeat buying, not by clicks alone. Clicks can reward recommending the same product that will be returned again.