Should Shopify product recommendations appear on the product page, in the cart, or both?
A shopper looking at a product page is still deciding what they want. A shopper in the cart has made at least one decision already.
That difference should shape your recommendations.
Putting the same four products in both places is easy. It also gives shoppers little new information. A better setup uses the product page for product-specific discovery, then uses the cart for additions that fit the order already in progress.
Affina is being built around that distinction. It will provide AI product recommendations on product and cart pages, learned for each store. It will also measure recommendations against a real 10% holdout, rather than treating clicks as proof that the recommendations helped.
Affina is pre-launch. It is not available to install today.
Start with the shopper's question
On a product page, the shopper's question is often: “Is this the right item?”
Recommendations can help answer it in two ways:
- Related products offer alternatives or nearby choices. Someone viewing a black rain jacket might want another jacket style.
- Complementary products help complete the use of the item. That same shopper may need waterproof trousers or a care product.
Shopify supports both related and complementary recommendations. Merchants can also manually choose recommendations for each source product.
The cart has a different question: “What else do I need before I check out?”
By then, the store knows more. It knows the exact items in the order. A cart recommendation should respond to that order, not repeat the browsing experience from the product page.
Say your store sells coffee equipment. On a grinder product page, a related recommendation might be another grinder. A complementary recommendation might be beans, a scale, or cleaning tablets.
Once the shopper adds the grinder, showing that grinder again in the cart makes no sense. Showing cleaning tablets may still make sense. Showing a second grinder probably does not.
Put product-specific suggestions on the product page
Shopify's Search & Discovery guidance is mainly built around product-page recommendations. Compatible themes add a complementary-products or product-recommendations section to the product template.
Shopify recommends placing complementary items near the top of the product page, around the product-information area and main image. That placement makes practical sense. The shopper can see the main item and its useful add-ons together, before they have moved on.
Keep the first set small.
Shopify's theme guidance suggests showing two or three complementary products by default. Its Product Recommendations API can return up to ten, but a page does not need to display every result it receives.
A long row of products asks the shopper to start browsing again. Two or three clear choices are easier to scan.
Product-page recommendations also have a useful built-in constraint in Shopify: they exclude products already in the visitor's cart. They also exclude out-of-stock products, zero-price products, gift cards, and products not published to the Online Store channel.
You can check your current product-page setup in under a minute:
- Open a product page on your phone.
- Find the recommendations section.
- Add one recommended item to the cart.
- Go back to the product page and refresh it.
- Check whether the item still appears, and whether the section is near the product details or buried below the page.
If shoppers need to scroll past reviews, policy text, and several collection rows to find the recommendations, the placement is unlikely to help them make a product decision.
Use the cart for order-aware additions
The cart is a better place for broader recommendations because the shopper has a basket, not just a single product view.
Shopify's checkout guidance says shoppers expect to edit orders in the cart. It also says checkout and order-confirmation offers should be limited and targeted, so they do not interrupt the original transaction.
That puts the cart in a useful middle ground. The shopper is still shopping, but the store now has enough context to make a more informed suggestion.
For the coffee store, a cart with beans and a pour-over dripper might suggest filters. A cart with a grinder and espresso beans might suggest a dosing cup. The right suggestion depends on the combination, not only on the most recently viewed product.
This is where a learned, per-store system matters. A recommendation model should not assume that every store's catalogue has the same relationships. A clothing store may see useful combinations by season or outfit. A skincare store may see routines. A parts store may see compatibility or replacement patterns.
The placement still needs restraint. Cart recommendations should help finish an order, not turn the cart into another category page.
A simple rule is this: if the recommendation needs a long explanation to make sense, it probably belongs on the product page. If it is an obvious useful addition to the current basket, the cart is a good candidate.
Use both, but don't repeat yourself
Shopify's merchandising guidance allows recommendation sections on either product pages or cart pages. That does not mean both placements should show the same catalogue row.
Give each placement a job:
- Product page: help a shopper compare, discover, or complete the item they are considering.
- Cart: help a shopper complete the order they have already started.
This avoids a common weak setup: the same “You may also like” products appearing everywhere, regardless of what is in the cart.
It also gives you a cleaner way to judge what happened. If a shopper adds a suggested item from the product page, that is product discovery. If they add it from the cart, that is order completion. Those are different behaviours and may deserve different recommendation logic.
Before adding a cart section, inspect your current flow. Add a common product to your cart and ask three plain questions:
- Does the cart already show recommendations?
- Are any of them already in the cart?
- Would a shopper understand why each suggested item belongs with this order?
If the answer to the last question is no, more recommendation slots will not fix it.
Don't measure clicks alone
A clicked recommendation is not necessarily a helpful recommendation.
Some shoppers would have bought the item anyway. Others may click around and leave. A recommendation can also shift an item from one place in the store to another without changing the final order.
That is why Affina is being designed with a real 10% holdout.
In plain terms, a holdout is a comparison group. Most eligible traffic sees recommendations. A smaller, randomly assigned group does not. You then compare outcomes between the two groups over the same period.
The holdout is not there to make reports look more cautious. It is there to answer the question that matters: did showing recommendations change what shoppers did?
This is especially important when you use both placements. Without a holdout, it is easy to count every recommendation click from the product page and cart, then assume both sections added value. With a holdout, you can look for the difference caused by recommendations as a whole.
A product page, cart, or both is not a permanent decision. It is a store design choice that should be tested against your own catalogue and customers.
For now, review your existing product and cart recommendations. Give each placement one clear job. Remove duplicates. Then decide what you would need to measure before calling either section a success.