Conclusion: virtual try-on usually increases conversion, but almost every published lift number is unverifiable, tracing to one vendor's press release or a single brand's blog rather than a controlled test. Shoppers who use a try-on do buy more often: on one live Ello store over the last 30 days (Sep 29 to Oct 28, 2026), 15.7% of try-on shoppers bought (17 of 108) versus 5.8% of everyone else (128 of 2,215), and of 111 shoppers who got a try-on, 45 added that item to cart and 17 bought it (15.3%). But try-on users self-select as higher-intent, so that gap is a comparison, not a measured lift. The only number that isolates what the try-on caused is a randomized holdout, where some shoppers never see the button; a tried-on-vs-didn't comparison measures who the shoppers are, not what the try-on did (our data; strict line-level attribution; one store, results vary).
Most quoted figures trace to one vendor's press release or a single brand's blog, not a controlled test. A number nobody can reproduce is marketing, not evidence. A shopper who opens a try-on already wanted the item more than one who scrolled past, so a tried-on-vs-didn't gap measures who they are, not what the try-on did.
| How you read it | The number (our data) | What it proves |
|---|---|---|
| Try-on shoppers who bought the item they tried | 15.3% (17 of 111) | They engaged and converted |
| Try-on vs no try-on, same store | 15.7% vs 5.8% | A comparison, not a cause |
| A randomized holdout | not yet run on this store | What the try-on caused |
Nearly 3x looks like proof. It isn't. Strip out the fact that try-on users self-selected and the honest answer is: we don't know the causal number until we run a holdout.
Usually, but most published lift numbers cannot be verified. On real stores, shoppers who use a try-on buy more often than those who don't, partly because they already wanted the item. A holdout test is the only way to isolate the effect.
There is no honest single number. On one live Ello store, 15.7% of try-on shoppers bought versus 5.8% of everyone else in the same 30 days (our data), but that gap is a comparison, not a measured lift. Distrust any vendor quoting a fixed percentage.
Run a holdout: randomly hide the try-on from some shoppers and compare conversion between the two groups over the same period. A tried-on-vs-didn't comparison is biased, because try-on users self-select as higher-intent.
Most trace to a single vendor's press release or one brand's blog, with no control group or published method. If a number cannot be reproduced, treat it as a claim, not evidence.