Size Recommendation Tools: When They Cut Bracketing and When They Do Not
Fit recommendation tools promise to end bracketing, and sometimes they do. The difference is data: tools fed on real kept-size history help, tools fed on brand charts alone mostly decorate the product page.
Every apparel retailer has been pitched the quiz: answer three questions, get your size, never return again. The results in practice split sharply. Some stores see multi-size orders fall by a fifth. Others see no change at all and a new line item on the invoice.
What the working versions have in common
The tools that move bracketing learn from outcomes. They know that shoppers who bought this dress in a 10 and kept it share a pattern, and that the 12 came back. Kept-size data is the one dataset that describes fit as experienced, not fit as designed.
They also speak in confidence, not verdicts. A recommendation that says most shoppers like you kept the medium, with an honest note when the data is thin, gets followed. A single confident answer on a product with eleven orders behind it gets ignored, correctly.
Why chart-driven quizzes fail
A quiz that maps height and weight to the brand's chart adds a layer of theater over the same numbers the shopper already distrusted. If the chart runs small, the quiz recommends small-running sizes with extra confidence, which is worse than no recommendation: the shopper orders one size, it fails, and the return is angrier.
Shoppers also game quizzes. Anyone who has answered 'between sizes' and been handed the smaller option learns to adjust answers until the tool agrees with the size she wanted. The session looks engaged; the bracket happens anyway.
Where tools cannot help at all
Some bracketing is not about the shopper's size. It is about preference: rise height, compression feel, how a shoulder sits. No quiz resolves taste. Tools also cannot fix inconsistent product, where the same size varies between production runs; they will recommend a size that was right last month.
Gift buying is the other blind spot. The buyer does not know the recipient's measurements and is bracketing on purpose as a service to the recipient. A recommendation tool aimed at the buyer answers a question she never asked.
How to evaluate one before you commit
Run it on one category with high bracketing and hold the rest as control. Measure multi-size order share and kept-size accuracy, not click-through on the widget. Vendors will show engagement numbers; engagement is not the outcome you are buying.
Ask what the model trains on and what happens with a new product with no history. The honest answers are kept-size data and a cold-start fallback to category patterns. If the answer is the size chart and your answers, you are buying a chart with better graphic design.
Related reading
These earlier guides go deeper on connected questions: Tall Sizing Bracketing: When Length, Not Width, Drives the Return, Maternity Bracketing: Sizing for a Body That Changes Every Month, and Workwear Bracketing Between Seasons: Sizing for Layers and Commutes.
Frequently asked questions
Do size recommendation tools actually reduce returns?
They reduce bracketing-driven returns in categories with enough order history, because they recommend the size shoppers actually kept. They do little for preference returns, quality returns, or products with thin sales history.
What data should a fit tool use?
Kept-size outcomes from orders and returns are the strongest signal. Brand size charts are the weakest, because they describe intended fit, and shoppers bracket precisely when intention and experience disagree.
Why do shoppers ignore the recommended size?
Usually because a past recommendation failed, or the tool gave a confident answer with no visible basis. Showing the basis, such as how many similar shoppers kept the size, restores a lot of trust.
Should a small brand buy a fit tool?
Only after fixing the free things: garment measurements, fit notes, and kept-size patterns in your own returns. A tool amplifies good fit data. It does not substitute for it.