Conversion · Product bundles

Product bundles: when a higher order value leaves less contribution

By Mirow · October 11, 2026 · 6 min read

Price a product bundle against the purchases it may replace. A bundle can raise average order value while leaving less contribution after product costs, fulfillment and media. The key question is how much it adds when someone buys an extra item, compared with how much it gives away when someone would already have bought both.

Start with one useful combination: products that solve a shared customer task. Then compare the single-product order, the full-price pair and the proposed bundle. A bundle's sales total cannot tell you which purchases it added.

Work out what the bundle replaces

For a fixed bundle containing one item A and one item B, there are several possible outcomes. An A buyer adds B. A shopper who wanted both takes the lower bundle price. Someone who would have left buys the set. Another shopper finds the larger commitment too expensive and leaves.

Those paths have different economics. Pull recent order IDs and quantities to see how often A and B already appear together, alongside single-item orders and other baskets. Separate distinct variants where compatibility matters. Existing co-purchase is a warning that the discount could subsidize demand you already have; it does not reveal how any particular customer will respond.

Shopify provides bundle and component sales reports, including a comparison of components sold in bundles versus individually. Those reports describe purchases. They do not show what each buyer would have purchased without the offer. Shopify sales reports.

Price three baskets before choosing a discount

Here is a hypothetical planning example in USD, not Mirow client data. Item A sells for $60 and item B for $30. The proposed one-of-each bundle is $75. Customers pay no shipping; tax is excluded. Assume no refunds, returns or cancellations in this simplified example.

The variable costs below include the product, payment fees, packaging, pick-and-pack labor and outbound carrier charges. Fee amounts are illustrative allowances, not a processor's pricing. The bundle assumes a $1 packing/fulfillment saving versus the full-price pair; that saving must be measured in a real operation. No other variable costs are assumed. Fixed overhead remains unpaid.

Scroll horizontally to see all columns.

Moving an A-only order to the bundle adds $15 of sales but only $3 of contribution before media: $32 − $29. Replacing a full-price pair with the bundle loses $14: $32 − $46. Both comparisons use the prices and cost assumptions shown above.

That means one discounted pair needs almost five A-only buyers to add B just to offset its $14 loss: $14 ÷ $3 = 4.67. Five such upgrades contribute $15. This is a planning hurdle under these assumptions; it excludes changes in order count, acquisition cost and future purchases.

A higher order value can hide a weaker result

Now compare two hypothetical outcomes for one test period. Each has 5,000 assigned eligible visitors, 100 completed orders and $1,500 in allocated media spend. The equal totals isolate the basket-mix mechanism; they are not a sample-size recommendation or a forecast that conversion stays flat.

Scroll horizontally to see all columns.

Average order value here means product sales after discounts divided by completed orders. Contribution after media is those sales minus the stated variable costs and media spend; its margin uses product sales as the denominator. All values are modeled, before fixed overhead.

The bundle outcome raises order value by 2.3% but reduces contribution by $190, or 10.9%. Per 1,000 assigned visitors, contribution falls from $348 to $310. Looking only at the $3,750 of bundle sales would miss that deterioration.

One assumed explanation is that 20 existing pair purchases take the discount, losing 20 × $14 = $280, while 30 A-only purchases add B, gaining 30 × $3 = $90. The net is −$190. Those switches explain this invented scenario. A real order export cannot identify the same customer's unobserved alternative purchase.

Check stock, checkout and returns

A workable price also needs a workable product. Before sending paid traffic, check the following with the people responsible for inventory and fulfillment:

• Component stock. For tracked inventory with overselling disabled, Shopify calculates bundle availability from the limiting component quantity. With 150 A units and 40 B units, a one-of-each set can support at most 40 bundles before other orders consume stock. A profitable single-product offer may also need those units. Shopify bundle inventory.
• Checkout price. Test the bundle with each permitted promotion. In Shopify Bundles, product-specific discounts need to include the parent bundle product; selecting only its components does not apply that discount. Component price changes also do not automatically update the bundle price. Recheck the economics after either changes. Shopify Bundles pricing and discounts.
• Partial returns. Check component-level return eligibility and refund allocation. Shopify's self-serve return rules can allow eligible components to be returned. Your comparison must deduct actual refunds once, then include unrecovered product, handling and shipping costs without double-counting recoverable inventory. Shopify bundle returns.

Test the offer across all eligible visitors

Use a concurrent random split where your testing setup supports it. Keep assignment consistent for repeat visits: one group can buy the existing products; the other can also choose the bundle. Keep the acquisition campaigns, standalone prices and other promotions consistent. Include everyone assigned to each group, including people who buy nothing. Comparing bundle buyers with single-item buyers selects different customer choices and cannot isolate the offer's effect.

Before launch, set an affordable spend cap, the smallest contribution improvement worth the work, and a review window that allows purchases and relevant returns to appear. Determine the sample requirement from your own baseline and variation. If available traffic cannot resolve a useful difference, treat the result as inconclusive and keep the commitment small.

Compare contribution dollars and contribution per assigned visitor, with margin, order count, component depletion and return costs alongside. Use a documented media-cost allocation for the groups; if it is unreliable, report contribution before media and show total spend separately. If variable costs cannot be reconciled, contribution is unavailable. Do not substitute order value for the missing costs.

If the problem is discount depth on an unchanged basket, use the free-shipping versus percentage-discount comparison. For this bundle decision, build the three-basket cost table first, then count how often the full-price pair already sells. Bring those two pieces of evidence into the next weekly budget review before committing production or media spend.