How can product recommendations increase order value?

Product recommendations raise order value when they offer a genuine complement where the shopper can still add it. Start with manual rules for bestsellers, one clearly defined block per placement and a control group. An expensive recommendation engine should follow a demonstrated need, rather than substitute for product knowledge.

The benchmark

Baymard distinguishes alternatives, which help people choose, from complements, which help complete the purchase. In its 2014 study of 50 large stores, only 42% offered both on product pages. Its 2021 cart research found irrelevant suggestions on 52% of desktop sites; a single doubtful recommendation could cause participants to ignore the rest. These are dated research samples, not current market-wide prevalence estimates.

The source's October 2026 Shopify documentation snapshot allows up to 10 related and 10 complementary recommendations per product. New products without purchase history can receive collection-based suggestions that are too broad. The source says description-based automatic analysis was limited to English-language stores; verify current language and account support rather than applying that limitation universally.

Placement can matter as much as selection. A baby-products brand in Le Club showed its “Complete your order” block only at the bottom of product pages, while roughly 65% of visitors viewed no product page at all. A book accompanying its flagship item was absent from recommendations, although one customer in 10 bought it spontaneously. The proposed cart test had not yet run. That is a diagnosis and test idea, not a measured uplift.

The rule

Manually define pairings for the 20–50 products responsible for most sales, then consider automation for the long tail. Choose the placement before the tool. Cart alternatives can reopen a completed decision; see upsell versus cross-sell placement.

Explain the reason: “Wear with,” “Compatible refill” or “Required batteries.” Baymard's research supports labels that help people understand the suggestion.

Keep the selection manageable and optional. In a field audit of an intimate-health brand, the authors completed a sensitive diagnostic only to receive a €542.30 routine behind one “Add everything” action. The audit criticized the abrupt jump from personal disclosure to a large bill. Let shoppers select or deselect each item and see the total update. The amount is the source's observed euro example, not a benchmark basket.

The steps

1. Find useful pairs. Inspect 12 months of orders for products bought with bestsellers, then validate that the association represents a real complement. Record contribution: a cheap accessory can raise order value but reduce profit if it fails to cover picking and packing. Use the order-margin method.
2. Write manual rules. In Search & Discovery, configure product recommendations and decide whether manual choices replace or supplement automatic suggestions. Confirm the current controls and whether the desired block uses those rules.
3. Inspect the component. Use a complementary-products block where the theme supports it. For cart recommendations, check the API request's intent=complementary. A fragrance brand found its cart carousel always showed the same list, even recommending the bottle already in the cart. A styled carousel is not proof of contextual logic.
4. Instrument placements. The source describes publishing a custom theme event with Shopify.analytics.publish, then passing appropriate data through a custom pixel to GA4 as view_item_list and select_item. Give each placement a stable item_list_name, such as product_recommendations, cart_recommendations or confirmation_recommendations, and preserve the list context on later cart additions. Check existing theme/app tracking first to avoid duplicates and honor applicable privacy settings.
5. Measure and compare. For each placement, inspect click rate, add-to-cart rate, average order value and contribution from orders including suggested items. Then test presence against absence using a suitable Shopify experiment, with revenue and profit outcomes alongside interactions.
6. Maintain the rules. Review unclicked pairs monthly, allowing for exposure volume. The source says native complements exclude unavailable and already-carted products; verify the actual behavior of your version, custom lists and third-party apps. Remove irrelevant or incompatible pairings promptly.

The pitfall

An app's “revenue generated by recommendations” commonly counts orders after someone clicked and bought a suggested product, including purchases that would have happened anyway. More exposure can make this attributed figure grow without improving the business.

Hide the block from a randomized control group and compare revenue per visit and contribution. With insufficient traffic, matched-period comparisons can provide an estimate, with explicit seasonal and assortment limitations.

What to tell your team

“The cart block gets clicks from 7% of visitors, but we do not know its incremental revenue. Before paying for an engine, we will define rules for 30 bestsellers and compare six weeks of revenue per visit against a control without the block.” These figures are fictional.

For related placements, see the confirmation page and on-site search.

Sources

• Baymard: alternatives and complements, 2014.
• cart cross-sells, 2021.
• Shopify: Search & Discovery recommendations.
• recommendations API.
• publishing pixel events.
• GA4: ecommerce measurement.