Variant Selector on Product Cards A/B Test Example

Put this test on your list when listing data shows shoppers moving back and forth between products to find the right color or size: quick returns from product pages, session recordings of detail pages opened one after another, or support tickets about sizes and colors. If the selector works, shoppers choose on the card, land on the product page with that variant already selected and reach a purchase in fewer steps. If you cannot show stock status reliably for each variant on the card, fix the stock data first; an out-of-stock option should never look selectable.

Purchase conversion rate

Compare purchases among eligible visitors assigned to the variant-selector experience.

Also track

• Listing-to-product click-through rateCheck whether visits to product pages change.
• Return-to-product-page rateReview rapid returns as a possible selection issue; the return alone does not explain why it happened.
• Return rateTrack whether selecting the wrong variant increases returns.
• Page load timeMake sure additional images do not slow the page.

• SelectorAdd a predefined variant selector to eligible product cards.
• BehaviorPredefine the available options and image-preview behavior, then keep them fixed.
• Keep constantKeep card size, product order and other card details unchanged.
• Selection carryoverCarry the selected variant through to the product page.
• MobileCheck touch targets; test option count or preview behavior separately.

1. 1If listing-to-product click-through rate rises while purchase conversion rate stays flat, the selector may be drawing interest without making the decision easier. A purchase lift is not enough to ship if return rate climbs or page load time gets worse. Check returns again once the test orders are past their return window.
2. 2Do not generalize a result from one category to categories without variants; if you plan to use the same selector in another category that has variants, validate it there. Review mobile on its own and confirm touch targets stay above the accessibility minimum, since cramped controls can show up later as wrong selections and returns.
3. 3Purchases are rarer than clicks, so before reading a small or unclear gap, enter the current purchase rate and the lift you care about into the sample size calculator to see the sample each variant needs. If the test did not reach that sample, record the result as unresolved by this test rather than as no effect. Do not change the displayed options or the preview behavior while the test runs.