Do virtual try-on tools actually cut bracketing? What the pilots show

September 28, 2026

Virtual try-on has been pitched as the end of bracketing for a decade. Point your phone at yourself, see the garment on your body, order the right size the first time. The pilots tell a more complicated story: real reductions in bracketing, but smaller than promised, concentrated in specific categories, and highly dependent on implementation details the vendors gloss over.

This post is a survey of what the pilots actually show, not what the marketing claims. If you are evaluating a try-on vendor, these are the questions that separate a useful pilot from an expensive demo.

What the pilots actually show

Across published pilots, the pattern is consistent: a modest reduction in multi-size ordering, typically in the single digits to low teens as a percentage, concentrated in fitted categories like denim, tailored suiting, and structured dresses. Loose-fit categories show little to no effect, which makes sense: when fit barely matters, seeing the garment on an avatar adds little information.

Return rates move less than bracketing rates. Shoppers who use try-on and still order two sizes tend to keep at similar rates as before, which suggests the tool helps at the margin: it converts some would-be bracketers into single-size buyers, but it does not fix fit for the shoppers who were always going to struggle.

Why the effect is smaller than promised

The core problem is the avatar. Phone-based body measurement is approximate, fabric drape is simulated rather than real, and the rendering flatters: garments look better on the avatar than they fit in reality. Shoppers sense this quickly. Trust in the tool erodes after the first order where the avatar said medium and the mirror said otherwise.

There is also a selection effect. The shoppers most motivated to use try-on are the ones already anxious about fit, who are also the most likely to bracket as insurance. The tool reaches the right audience but cannot fully overcome the anxiety that brought them there. A shopper who does not trust the size chart will not trust the avatar either.

What makes the difference in the pilots that work

The pilots that move the number share traits. They pair the visual try-on with a size recommendation grounded in real data: your returns history, the garment's measured specs, and the shopper's kept-size history. The avatar is the engaging surface; the recommendation engine is what actually changes the order. Vendors that sell only the visualization consistently underperform vendors that sell visualization plus data.

Placement matters too. Try-on buried three taps deep in the product page gets used by almost nobody. The pilots with real adoption put the entry point next to the size selector, at the exact moment of size uncertainty, with copy that promises a recommendation rather than a gimmick. Framing it as a fitting tool rather than a toy changes who tries it.

The honest evaluation framework

If you pilot try-on, measure bracketing rate among users versus a matched control group, not versus site average. Users self-select, so the comparison has to be apples to apples. Track kept-size accuracy: of the shoppers who followed the recommendation, how many kept without a size exchange? That is the metric the vendor will not volunteer.

And price the pilot against the alternative. A size recommender built on your own returns data, or even better fit content on the product page, often delivers a comparable bracketing reduction at a fraction of the cost. Virtual try-on is not useless, but it is rarely the highest-ROI fit investment. Spend on the data layer first, the visualization second, and you will know exactly what the avatar is adding.