How do I implement product recommendations?

Start with platform-native recommendations — Shopify’s native engine or WooCommerce plugins work well for most stores. Product recommendations placed on the cart page typically generate 10-30% of total recommendation-driven revenue, more than homepage or PDP placement alone.

The recommendation placement hierarchy

Where you place recommendations matters as much as what you recommend. Based on conversion data across e-commerce stores, the highest-converting placements are:

1. Cart page (“You might also like” / “Frequently bought together”) — Users who have added to cart have already made a partial commitment. Basket expansion here adds to an existing order without needing to restart the purchase decision. Average basket-size lift from cart recommendations: 15-25%.

2. Product detail page (“Customers also bought”) — The highest-traffic recommendation placement, but lower conversion per click because users are still in evaluation mode. Valuable for breadth of product discovery.

3. Post-checkout thank you page — An underused placement. A customer who just converted has peak trust in your brand. Cross-sell recommendations here have no friction cost to the completed order and commonly drive immediate second purchases or wishlist saves.

4. Homepage (“Based on your browsing” / “Recently viewed”) — Only valuable for returning visitors. Generic “bestsellers” or “new arrivals” on the homepage are static merchandising, not personalization.

5. Category pages — Useful for surfacing related categories or trending items within a specific range. Lower conversion than PDP or cart placements but meaningful for discovery.

Implementation by platform

Shopify: Shopify’s native product recommendations API powers the default “You may also like” section. It’s algorithm-driven and decent, but limited in customization. For better control, apps like LimeSpot, Rebuy, or Nosto layer on top with more placement options, more recommendation types, and A/B testing.

Cost range: Free (native) to £200-500/month for full-featured platforms.

WooCommerce: WooCommerce includes basic “Related products” and “Customers also bought” out of the box. For AI-powered recommendations, plugins like Beeketing or Barilliance extend capability significantly.

Headless or custom: If you’re building custom, Algolia Recommend or Recombee provide API-driven recommendation engines that integrate with any frontend.

Choosing recommendation algorithm types

Different algorithms serve different goals:

• “Frequently bought together” — collaborative filtering based on cart co-occurrence. Best for bundle creation and basket expansion.
• “Customers who viewed this also viewed” — useful for product discovery within adjacent categories.
• “Similar products” — content-based similarity. Useful when collaborative filtering lacks data (new products, low-volume SKUs).
• “Trending now” — time-weighted popularity. Works well for seasonal or trend-driven categories.
• “Recently viewed” — pure behavioral, no algorithm. High click-through because it’s personally relevant.

Most stores should implement at least two algorithm types across different placements, because one approach doesn’t serve all contexts equally well.

Measuring recommendation performance

The correct metrics for product recommendations:

Click-through rate (CTR): What percentage of users who see the recommendation widget click it? Benchmark: 5-15% for well-placed, relevant recommendations.

Revenue influenced: Total revenue from orders where a recommendation was clicked, expressed as a percentage of total revenue. Target: 10-25%.

Uplift in average order value: Compare AOV for sessions where a recommendation was clicked versus sessions where it was not.

Avoid: “recommendation views” as a success metric — impressions without revenue impact don’t matter.

Common implementation mistakes

Recommending out-of-stock products. Few things frustrate converting customers more. Filter your recommendation feeds for in-stock items only, in real-time.

Recommending the same product the user just added. If someone adds a blue sweater to cart, showing them the same blue sweater as a recommendation is a waste. Exclude the current product from all recommendation contexts.

Slowing your page with heavy recommendation scripts. A recommendation widget that blocks rendering or adds significant load time will cost more in conversion from speed degradation than it gains from relevance. Lazy-load recommendation components below the fold.

Not testing placement. Run an A/B test on whether recommendations appear above or below the fold on your product page. Placement can affect CTR by 40-60%.

Start with the cart page — it’s the easiest win and has the clearest revenue attribution. Book a call to discuss a full recommendations strategy for your store and platform.