What a virtual fitting-room A/B test really measures
A well-built test answers one question: does the fitting room make people buy, and do those purchases stay bought.
WHAT TO MEASURE
Four numbers are enough: add-to-bag rate, conversion rate, 30-day return rate and average order value. The first three carry the promise. The fourth exposes the side effects.
HOW TO SPLIT TRAFFIC
Split on entry to the product page, 50/50, per visitor rather than per session. A returning shopper must see the same experience, or the return figure becomes unreadable.
HOW LONG
Returns land after delivery, not after checkout. Count the purchase cycle plus the return window: six to ten weeks on most catalogues.
A test that stops at checkout measures half the subject.
THE PITFALLS
Bracketing, ordering two sizes to keep one, reads as both a higher basket and a higher return rate. Isolate it, or it hides the gain. Seasonality moves both groups at once: it does not bias the comparison, but it makes absolute figures useless out of context.
WHAT A GOOD RESULT LOOKS LIKE
On a womenswear catalogue, a credible gain sits between +5 % and +15 % conversion on equipped pages, and 10 % to 30 % fewer returns on orders placed after a try-on. Above that, check your sample size before publishing it.