Order Confirmation Page Product Suggestion A/B Test

Customers who come back soon after ordering for related products, or support requests asking to add an item to an order already placed, suggest buyers still want complementary items after checkout. When the recommendations land, some buyers pick one and complete a second order separate from the first. If the requests are about adding items to the original order, look at an order-editing option first. Do not show guest buyers an account-creation prompt on this page during the same test; it is a separate variable, and you would not be able to tell the two effects apart.

Additional purchase rate

Measure completed additional orders per assigned order-confirmation page visitor over a fixed follow-up window.

Also track

• Value of additional ordersReport the value of completed orders that are separate from the original order.
• Order summary accessTrack whether users open the order summary after seeing recommendations; order details must remain visible.
• Order-related support requestsCompare questions about order status or tracking using the same denominator.
• Confirmation-page exit rateInterpret exits alongside access to order information; leaving after confirmation can be normal.

• ChangeAdd a related-product recommendation module to the order-confirmation page.
• Keep constantLeave the order summary, delivery details and confirmation content unchanged.
• BehaviorReview recommendation interactions alongside completed additional orders.
• Order informationDo not let the module obscure the order confirmation, number or delivery information.
• Separate testsTest placement, recommendation count and recommendation logic separately.

1. 1Clicks on recommendations or started orders do not decide the test; count only additional orders completed within the fixed window. Report the value of additional orders separately from the original order, and do not read confirmation-page exits as bad on their own, since leaving after an order is normal.
2. 2If order-related support requests or order summary access change, the module may be confusing buyers. Confirm the recommendation card can be dismissed, that the confirmation is reachable without getting past it, and that a purchase started from it does not suggest the original order still needs payment.
3. 3Treat the outcome as specific to the module, placement and product-selection logic you tested; a different recommendation algorithm or count is its own experiment. Validate it again in context before moving it to other post-purchase screens.
4. 4Should the difference stay unclear, note that this sample could not separate a change in additional orders, rather than reporting that the module had no effect. Once the follow-up window closes, enter each arm's confirmation-page visitors and buyers who completed another order into the A/B test significance calculator; it shows whether the gap is beyond chance. It does not test averages, so treat any difference in the value of additional orders with the same caution.