How to measure if virtual try-on actually lifts sales

Most try-on dashboards show you a big number: shoppers who used try-on converted far better than shoppers who didn’t. It’s tempting to take that as proof. It isn’t, and knowing why will save you from paying for a feature that only looks good in a report.

The problem with “users of try-on convert more”

Shoppers who click Try it on are not a random sample. They are already interested enough in the product to spend a minute on it. Many of them would have bought anyway.

Comparing them with everyone else mixes two things:

the effect of try-on, and
the fact that engaged shoppers convert more than casual ones.

This is selection bias, and it can make any feature look like a success, whether it changes behaviour or not.

The fix: a holdout group

The reliable way to measure a feature is to compare two groups that differ only in whether they could use it.

Split visitors at random. Most visitors see the Try it on button as usual. A small share, the holdout, never sees it.
Keep the split stable. A visitor should stay in the same group on every visit, otherwise the groups blur together.
Compare whole groups, not clickers. Measure conversion and revenue for all visitors in each group, including those who never tapped the button.

Because the groups were chosen at random, the only systematic difference between them is access to try-on. Any gap in conversion is the effect of try-on itself.

How big should the holdout be?

It’s a trade-off. A larger holdout gives a clear answer sooner, but more visitors miss out on try-on while you measure. A small holdout of around 10% is a common starting point for stores with steady traffic. Smaller stores may need longer to see a clear result.

Reading the result

Two numbers matter:

Lift: how much higher conversion (or revenue per visitor) is in the try-on group than in the holdout.
Confidence: how likely it is that the difference is real and not noise.

Early on, differences jump around. Don’t stop the test the first day it looks good. Wait until the result is marked significant, and check it against a full week or more so weekday and weekend shoppers are both included.

What else to watch

Conversion is not the only outcome. Try-on can also change:

Return rate. Shoppers who saw the product on themselves may keep it more often. Track the return rate of tried-on products against the rest.
Order value. Shoppers might add a second item they tried on.
Product mix. Some products may benefit far more than others; that tells you where to promote try-on.

How Mirava does it

Mirava has the holdout built in on the Growth and Scale plans. You choose the share of visitors in the holdout, the split stays stable per visitor, and orders are attributed through cart attributes, so no customer details are stored. The dashboard shows conversion with and without try-on, the lift, and whether it is significant.

If you’re still choosing a try-on app, read what AI virtual try-on is and what to check first.