Shopify product recommendations work best when related products help shoppers compare and complementary products help them complete a purchase; map each role to a product-page decision before adding a widget.
A recommendation row is not automatically a cross-sell. It is a merchandising decision about what the shopper should do next:
Shopify gives merchants both related and complementary recommendation intents, but the quality of the result depends on the product relationships, catalog data, theme placement, and measurement around them.[1] This workflow helps the Shopify ecommerce merchandiser make those decisions deliberately for fashion, beauty, and other physical-product catalogs.
Before opening Search & Discovery, write down what the shopper is trying to decide on the page.
| Shopper job | Recommendation role | Useful relationship | Example |
|---|---|---|---|
| Compare alternatives | Related | Similar product that could replace the current choice | Another blazer with a different cut or price |
| Complete a use case | Complementary | Product that works with the chosen item | A belt that completes an outfit |
| Build a routine | Complementary | Product that belongs to the same verified routine | A cleanser paired with a moisturizer |
| Recover from a poor fit | Related | Substitute with a meaningful difference | A different size system or silhouette, when the page explains the difference |
| Browse a collection | Neither by default | A collection, filter, or search path may be more useful | A shopper who has not chosen a product yet |
A related product should give the shopper a credible alternative, not merely another item from the same catalog. A complementary product should add utility without making the shopper question the primary choice.
For fashion, the distinction might be another dress in a different cut versus a matching belt. For beauty, it might be another cleanser for a different skin concern versus a moisturizer that belongs in the same routine. For a broader product catalog, it might be a replacement laptop sleeve versus a cable organizer that fits the selected laptop.
Create one row for each relationship you want a customer or teammate to understand. The map can live in a spreadsheet, product brief, or merchandising system that your team already maintains.
| Field | Question to answer |
|---|---|
| Primary product | Which exact product and variant starts the recommendation? |
| Shopper job | Is the customer comparing, completing, replacing, or learning? |
| Role | Is the recommendation related or complementary? |
| Relationship statement | Why does this item belong beside the primary product? |
| Compatibility | Does size, shade, fit, material, model, voltage, or routine order matter? |
| Availability | Can the recommended item be bought in the market and channel where it appears? |
| Evidence | What product data, support question, review language, or order pattern supports the pairing? |
| Placement | Product page, collection, cart, or another surface? |
| Fallback | What should the shopper see when the relationship is unavailable? |
| Owner | Who reviews the relationship when the catalog, price, or inventory changes? |
Write the relationship in customer language. “Pairs with the 13-inch model” is more useful than “same accessory collection.” “Adds warmth after cleansing” is more useful than “same skincare category.” A clear relationship statement also gives the copy, design, and support teams a way to spot a bad recommendation.
A product being in the same collection is a starting signal, not proof that it belongs in a recommendation row. Use the inputs your team already owns:
Label the evidence as observed, manually curated, or inferred. An observed co-purchase pattern can support a recommendation, but it does not explain whether the relationship is useful for every market, variant, or customer. A manually curated pairing can be right on launch day, but it needs an owner and a recheck condition.
If the relationship exposes a product-copy gap, compare the evidence with this product-description refresh workflow before changing the recommendation label.
Related products are alternatives to the product the shopper is viewing. Shopify describes them as similar products that help customers discover potential substitutes.[1]
Use this role when the shopper may still be deciding between:
The relationship should explain the difference. “Compare relaxed and tailored fits” gives the shopper a reason to open another product. “You may also like” is a useful theme label, but it does not replace the merchant’s responsibility to inspect what the row actually shows.
Shopify’s automatically generated related recommendations can use purchase history, similar product descriptions, and related collections, depending on the data available to the store.[1] Treat those inputs as signals to review, not as a guarantee that every returned item is a suitable substitute.
Complementary products are add-ons that work with the selected product rather than replace it. Shopify identifies complementary products as a separate recommendation intent and expects merchants to configure them manually.[1]
Use this role when the shopper has a clear primary choice and the additional product:
Do not use complementary recommendations as a disguised product bundle. A bundle may change price, inventory, fulfillment, or return handling; a recommendation only proposes a next product. If the relationship needs a combined price or an all-in-one inventory promise, evaluate it with the store’s bundle and operations workflow instead of changing the recommendation label.
A correct relationship can still disappear from the storefront or show the wrong set if the product and theme conditions are not met.
Shopify’s current recommendation guidance says that a recommended product must be active, published to the Online Store channel, priced above zero, and not a gift card or unlisted product; it also cannot already be in the visitor’s cart.[1] Related products can display when inventory is zero if the store allows continued selling, while complementary products require inventory above zero.[1]
Use those rules as a preflight, not as a substitute for a rendered-store test. Check the actual product, variant, market, price, availability, and theme output that a shopper will see.
The product page also needs theme support for the relevant recommendation surface. Shopify documents a related-products section and a complementary-products block, and notes that the complementary block is available only in compatible themes.[3][1] A recommendation configured in the app is not proof that a shopper can see it in the published template.
Before launch, inspect:
If a recommendation is not useful without a particular variant or market, do not let a generic row make the shopper infer compatibility. Use a precise relationship, a product-page explanation, or no recommendation.
The same pairing can be helpful in one location and distracting in another.
Place related products where a shopper is comparing the primary choice, and place complementary products after the page has made the main product understandable. Keep the recommendation heading explicit enough that the shopper can tell whether the row offers alternatives or add-ons.
For a fashion page, a related row might help someone compare two fits, while a complementary row might appear after size guidance with a matching accessory. For a beauty page, a related row can help compare formulas, while a complementary row can help build a routine only when the order and suitability are clear.
A shopper who has not selected a primary product may need navigation, filters, search, or collection merchandising rather than a product-specific recommendation. Do not force a product-page relationship into a collection surface just because the theme provides a carousel.
A cart or post-purchase offer should not reopen a decision the shopper has already made unless the change is intentional. Prioritize one clear, compatible add-on over a long list of alternatives. If the offer changes shipping, discount, tax, or fulfillment behavior, assign it to the team that owns that operation before publishing it.
Start with a product group where a wrong recommendation would be visible to the team and meaningful to customers. Do not begin by manually pairing the entire catalog.
For each primary product, record:
Use category-specific failure examples:
Keep to make every shown relationship defensible, understandable, and available as the decision, with to maximize the number of rows as supporting evidence.
Choose one primary metric before changing the recommendation surface. A practical default is the purchase rate for recommended products among sessions that viewed the recommendation surface, but the right metric depends on whether the recommendation is meant to save a comparison, add a compatible item, or support a later purchase.
Shopify’s Product recommendation conversions over time report describes sessions that viewed recommendations, sessions with clicks, sessions with recommended add-to-carts, and sessions with purchases.[2] The report also exposes click rate, add-to-cart rate, and purchase rate for the recommendation path.[2]
Read those measures as a funnel, not as proof that the recommendation caused the sale. Shopify says the report counts interactions within the session and does not include products added or purchased through a Buy Button or third-party app.[2] Compare the same surface, product group, traffic conditions, and time window when you evaluate a change.
Pair the main measure with risks to monitor:
Record the baseline, the change, the affected product group, and the conditions of the observation. If clicks rise but recommended-product purchases or customer questions worsen, the surface may be attracting attention without helping the decision. If the success measure is unchanged, keep the recommendation only when it solves a documented customer problem or improves discovery without adding operational cost.
Use this review brief after the first audit or after a catalog change:
| Finding | Action | Evidence to retain |
|---|---|---|
| The item is a clear substitute and the comparison is useful | Keep as related | Difference in fit, price, material, or use case |
| The item completes the chosen product and compatibility is clear | Keep as complementary | Relationship statement and availability check |
| The item is useful but the role is ambiguous | Improve | New heading, placement, or relationship mapping |
| The item is frequently unavailable or wrong for a variant | Remove or remap | Affected products, variants, and support/return signal |
| The row is generic and no customer job is documented | Remove | Reason the surface does not help the page decision |
| The surface is not visible in the published theme | Fix the template or hide the block | Theme, template, and mobile test result |
Recheck relationships after a meaningful catalog event: a product launch, price change, discontinued SKU, new market, material change, fit update, or theme release. A recommendation that was correct at launch can become misleading when the catalog changes around it.
Before enabling or expanding the surface, confirm:
Product recommendations earn their space when they make the next decision easier. Start with a small set of relationships you can explain, verify the published buying path, and expand only when the evidence and operational ownership support the next group.
[1] https://help.shopify.com/en/manual/online-store/search-and-discovery/product-recommendations — Shopify Help Center: Customize product recommendations [2] https://help.shopify.com/en/manual/reports-and-analytics/shopify-reports/report-types/default-reports/behaviour-reports — Shopify Help Center: Behavior reports [3] https://help.shopify.com/en/manual/online-store/themes/customizing-themes/common-customizations/add-product-recommendations — Shopify Help Center: Adding a related products section