Testing Product Recommendations Inside the Cart

Order data showing items used together being bought in separate orders, or support tickets asking which items are compatible, suggest that carts tend to reach checkout with a single item. If the suggestions fit, shoppers add a suggested item and still move on to checkout just as easily. If you already see shoppers dropping off between the cart and checkout, find the cause of that loss first; a recommendation module will not remove it. Also confirm before launch that prices and discount information on the recommendation cards match the product pages.

Average order value (AOV)

Compare revenue per completed order, and read this buyer-only average alongside revenue per visitor.

Also track

• Add-to-cart rateCheck whether the recommendations change add-to-cart behavior.
• Order completion rateTrack order completion among assigned cart visitors after recommendations are added.
• Revenue per visitor (RPV)Check whether any change in AOV also changes revenue per visitor.

• ChangeAdd a complementary-product recommendation module to the variant.
• Keep constantKeep checkout flow, prices and the main cart CTA unchanged.
• RelevanceRecommend only in-stock products that are genuinely related to items in the cart.
• InterpretationMonitor order completion and revenue per visitor alongside AOV.
• Separate testsTest recommendation count, order, placement and card design separately.

1. 1Do not count recommendation clicks as success. Decide on AOV, and read order completion rate and revenue per visitor (RPV) alongside it; if AOV rises while order completion falls, that does not justify keeping the module.
2. 2This finding is limited to this module in this cart placement; if you plan to use it elsewhere, such as on product pages, test it there again. In this test the module should not push the checkout CTA into the background or sit above it; recommendation count, order, placement and card design each need their own test.
3. 3AOV is an average over buyers only, so a few large orders can move it. Do not read a small or unclear difference as either a win or no effect. Test whether the add-to-cart rate gap is beyond chance with the A/B test significance calculator; it does not cover averages, so weigh the AOV gap with the same caution.