Bundles From Similar or Mixed Price Tiers A/B Test

When shoppers pick single items over the bundle, remove items from it, or give return reasons tied to its contents, the way bundles are put together may be costing sales. If composition is what matters, one bundle brings either more purchases or larger orders, and revenue per visitor reflects it. If bundles are being missed because of the price explanation or placement, look at those first. Before launch, fill both bundles with in-stock, good-quality, related products, and if you show savings, calculate them from actual standalone prices.

Revenue per visitor (RPV)

Compare revenue per visitor for the two defined bundle compositions.

Also track

• Bundle selection mixReport which bundle or single-item option each visitor selects.
• Return rateCompare returns over the same follow-up period and record bundle contents and return reasons.
• Average order valueCalculate per completed order, using the same revenue definition for discounts and cancellations.

• CompositionDefine one bundle made of similarly priced items and one mixed-price bundle.
• Product setKeep quantity and use case as similar as practical; record unavoidable product differences.
• PresentationKeep price explanation, placement and any savings calculation unchanged.
• FlexibilityKeep item-removal and single-item purchase rules the same.
• AudienceReview predefined buyer groups.

1. 1Let revenue per visitor decide; use the bundle selection mix to see whether revenue comes from bundles or single items, and average order value to see the effect per order. Keep the single-item option equally visible in both arms; a layout that pushes shoppers into the full bundle inflates revenue, so do not ship on that kind of gain. If revenue rises along with bundle-related returns, question the result.
2. 2Because the products differ, the result compares these two specific compositions; do not credit price range alone, and note unavoidable product differences in the report. Do not let the bundle's total price change along with its composition, or you cannot tell which one caused the difference. Read predefined buyer groups separately, and compare returns only after the follow-up period has run its full length in both arms.
3. 3A few large orders can move revenue per visitor, so treat it as a noisy metric. The A/B test significance calculator tests rates, not averages, so use it on the share of visitors who buy in each arm and treat the revenue gap with the same caution. If both stay unclear, report the result as inconclusive.