What size and fit information should apparel shoppers see?

For apparel where sizing is difficult, test an optional personal size recommendation beside clear garment measurements. Keep the chart available so a shopper can check the advice.

Gallino and Moreno (2018) randomly offered virtual fit information to online apparel customers. The publisher’s abstract reports more purchases and higher order value, alongside fewer returns and fewer orders containing multiple sizes of the same product.

The verified primary material establishes the direction of these results. The full numerical results were not recovered from the publisher, author record or institutional searches, so no conversion uplift is quoted here. Treat this as a reason to run a controlled test, rather than a forecast for your store.

A useful recommendation can answer “which size should I order?” before a shopper pays. That may reduce the need to use the delivery network as a fitting room. The study evaluates a virtual fitting service; a simple size quiz or a new chart still needs its own test.

The accessible primary record does not provide the arm sizes, effect estimates or uncertainty needed to reproduce the results. Availability was randomized; voluntary engagement was not. This finding does not establish that every fit tool will reduce returns, or that its software cost will pay for itself.

Pair “runs small”, “true to size” and “runs large” with the size purchased and relevant reviewer context. Let reviewers choose which body details they are comfortable sharing.

Wang, Ramachandran and Sheng (2021) studied a US outdoor retailer that added reviewer height, weight and purchased size to its fit-review system. Their analysis compared product histories with and without reviews containing both a fit opinion and its reference, using a model to construct the comparison. Shoppers were not assigned at random.

In the matched analysis of 333 products in each group, the estimated monthly product return rate was 5.3 percentage points lower after these combined reviews became available, relative to the model’s comparison. The table reports −0.0530 on a rate measured from 0 to 1, with a standard error of 0.0147. To picture the scale, imagine each compared product had exactly 1,000 orders in a month. A 5.3-point lower return rate would correspond to 53 fewer returns per product. Those equal order counts are hypothetical: the study averaged product-level monthly rates, so the estimate does not establish 53 fewer returns per 1,000 orders across the store.

Fit opinions without that reference were less consistently useful. The authors found that an opinion alone could help when enough reviewers agreed; context mattered especially when opinions were sparse or conflicting.

“Runs small” describes a relationship between a garment, a person and the size they tried. A size label alone leaves two of those pieces missing. Context helps the next shopper decide whether the same opinion applies to them.

The result is a model estimate from one retailer, not a randomized layout test. Products received contextual reviews at different times, so the conclusion depends on the model’s assumptions. The outcome averages product-level monthly return rates, rather than pooling every order; it tracks returns through November 2015 for purchases through September. A shorter return window produced a smaller estimate.

Review a fit tool after returns have settled, using sales customers keep and the full cost of serving them. Tool usage and placed orders are useful diagnostics, but neither proves better fit or higher profit.

Patel, Karlsson and Oghazi (2025) examined 496,365 ordered items from 75,707 customers at one Sweden-based fashion retailer. Customers chose whether to use its size finder. The adjusted analysis associated use with a slightly higher likelihood of an item being returned; a separate analysis also associated more use with higher later customer purchasing value.

This is a useful counterweight to assuming that a fit tool must lower returns. It does not establish that using the tool caused more returns or more valuable customers. The paper’s headline return percentage depends on an assumed baseline and an unclear model description, so it is not used here as an expected number of extra returns.

A shopper may buy because guidance makes them more confident, yet still return some of those purchases. Someone already unsure about fit may also be more likely to open the tool. A controlled test and a complete cost calculation help separate those explanations.

These are observational records from one retailer, with exchanges and categories without the tool excluded. The paper reports inconsistent language around probability, odds and the estimation method, which limits interpretation of its small return effect. Later purchasing value is not the same as profit after returns. Do not use this study to forecast a causal penalty or gain.