A practical guide to placing recommendations in search and collection pages

In short, when you add recommendations into search results and collections, three things matter: 1) do not drift too far from the user’s original intent, 2) limit recommendation exposure to roughly the top 20–30% of the list, and 3) design around “what you want them to see next in your inventory”, not just current bestsellers. If you simply sort everything by “popularity”, you’ll keep surfacing sold‑out items or low‑margin products and lose out on both revenue and inventory turns. In real stores it becomes much easier to design if you treat search and collections as having different roles. Below, everything is broken down at a level you can implement in Shopify right away. Note: this article covers general operational know‑how only; it does not depend on any specific theme or app. For concrete UI steps, follow the documentation of your theme or apps.

Treat search results and collection lists as different surfaces

The first thing to keep in mind is that even though “search” and “collections” both look like product lists, users come to them with very different goals. Search is mainly used by people who are “looking for something specific”. Collections are used more by people who “want to browse and compare”. This difference directly affects what kind of recommendations you should mix into each.

In search results, the main actors must be products that match the query. If someone searches for “black T‑shirt” and the top row is filled with recommended coats, they will just leave. A collection, on the other hand, is a category list, so it is a natural place to do merchandising that reflects bestsellers, new arrivals, and inventory priorities. Even though both are “recommendations”, your design should shift: prioritize relevance in search, and prioritize what you want to sell in collections.

A common failure pattern is reusing exactly the same “sort by popularity + recommendation banners” in both search results and collections. If you do this, search results get cluttered with items that do not fit the intent, while collections get dominated by a handful of popular products. As a result, items with surplus stock or higher margins are barely seen, and inventory turns get worse.

Recommendations in search results: what to place where

On the search results page, the basic structure should be: “first show items that match the query”, and “on top of that, add recommendations that support users who are hesitating”. The placement of recommendations should be constrained: a small block at the top, or somewhere in the middle or bottom of the list. If you start the very first row with recommendations that are weakly related to the query, click‑through rates tend to fall.

In practical operations, the following patterns are relatively easy to work with.

• Display a single row (for example, 4 products) of “Recently viewed” or “Recommended from your browsing history” above the search results
• Limit the first one or two rows of the results list to only products that match the query
• Only when there are few search results (for example, fewer than 5 products), show “Popular items in this category” at the bottom of the page

In an offline store, search is similar to “a customer has told you the product name they want”. In that situation, staff would never walk them over to a completely different shelf first. They would show the relevant shelf, and quietly place related items or higher‑end models around it. If you design search recommendations with that image in mind, you are less likely to get it wrong.

As a numerical guideline, keep “recommended” slots on the search results page to around 20–30% of all displayed items so you do not upset the balance. For example, on a search results page showing 24 products, you might limit recommendations to around 4 “Recently viewed” items plus 2 “Popular in this category” items, and dedicate the rest of the slots to actual search results.

Effective recommendation patterns and ordering in collections

The collection page is the core stage for merchandising. Here you want it to do more than “sell more of what already sells”. You also want it to “bring priority inventory to the front” and “get shoppers to notice new arrivals”. When placing recommendations, decide up front which logic will fill which rows; this makes ongoing operations much more stable.

One commonly used ordering rule looks like this:

• Rows 1–2: Pin the collection’s bestsellers and high‑margin products
• Rows 3–4: Prioritize items with lots of stock or those you want to push strategically
• Row 5 and later: Use the standard sort order (for example, newest first or bestsellers)

The key is to mix “fixed slots” with “automatic sorting slots”. If you make everything popularity‑based, out‑of‑stock or size‑broken items tend to linger at the top forever. But if you fix everything manually, the update workload gets heavy and things get out of date. If you only pin the top 8–12 products and leave the rest to automatic sorting, it becomes much easier to balance operational effort and results.

A frequent pitfall is “prioritizing sale items so hard that full‑price sales collapse”. Instead of filling every row with sale merchandise, keep contrast: for example, row 1 is full‑price bestsellers, row 2 is sale items. This way you can protect overall margins while still clearing inventory.

Choosing recommendation logic: popularity, similarity, and complements

Even when the block is labeled “Recommended”, its role changes depending on the underlying logic. When embedding recommendations into search results or collections, it helps to consciously use three types: “popularity” (top sellers), “similar items”, and “complements” (items often bought together).

Within search results, the baseline is “products close to the query + popularity”. For example, if someone searches for “black sneakers”, you would list black sneakers ordered by popularity or review count, and then lightly add complementary logic around them, such as “socks often bought together”.

On collection pages you can take a more aggressive approach. Similarity logic, which surfaces items with a similar style within the same category, is good for increasing browsing depth. Complement logic that shows “items often bought together in this collection” ties directly into bundle sales and higher average order value.

What you want to avoid is making everything “popular across the entire store”. For example, if you do that, men’s products that are popular store‑wide can slip into a ladies’ collection, breaking the context and driving people away. When you configure logic, deliberately base it only on data that fits “the context of this surface” (search results or this specific collection).

How many recommendation blocks, and where: focus on the top 20–30%

A common struggle with recommendations is “how many items to show, and where to put them”. Balancing user experience and revenue, it is realistic to limit recommendation slots to about 20–30% of each list page. Show too many and it starts to feel like stealth advertising; show too few and the impact becomes hard to see.

The following basic layout patterns tend to be easy to manage:

• Row 1: Products that match the page’s core purpose (query‑matched items for search, bestsellers for a collection)
• Rows 2–3: Recommendation rows (driven by recommendation logic)
• Row 4 and later: Standard sort order (newest, popular, price, etc.)

A typical real‑world mistake is to fill the very top of the page entirely with recommended items “to measure the impact of recommendations”. If you do this, the list ends up full of products that do not match the search intent or category expectations, and bounce rate goes up. It is safer to first test inserting recommendations into rows 2–3, then look at changes in click‑through and revenue before increasing or decreasing exposure.

Also remember that on smartphones, there are fewer products per row. If you increase the number of recommendation rows using the same “row count” mindset as on desktop, recommendations can end up occupying almost the entire page in practice. Actually scroll through your store on a phone and check “within the first 1–2 screens, how much space is taken up by recommendation blocks”.

How to put this into practice with RecoBoost

When you use RecoBoost, first decide your rules for “placement and role” as described above. Then, on the search results page, insert a single row of “query‑aware + browsing‑history‑based recommendations”, and on collection pages, use a design where you “pin the top few items + add 2–3 rows of popularity/stock‑aware recommendations within the collection”. With RecoBoost’s scenario and pinning features, it is easy to formalize hybrid setups like “rows 1–2 are manually pinned, and everything after that is recommendation logic” for each store. Combine these with your theme and other apps, and gradually fine‑tune the balance for your own store to keep operational effort low.

When placing recommendations in search results and collections, the key points are: 1) be clear about the page’s role (finding vs. comparing), 2) design recommendation slots so they are limited to the top 20–30% of the list, and 3) mix in logic that considers inventory and margin, not just bestsellers. Combine fixed placement with automatic recommendations, run gradual A/B tests, and find the balance that works for your own store.