Recommendation engine

What is a recommendation engine?

A recommendation engine analyzes data on customer behavior and uses algorithms to suggest products that an individual customer is likely to be interested in. You’ve probably seen this in the “You might also like…” or “Customers who bought this also bought…” sections on product pages and in emails.

Recommendation engines account for a large portion of revenue in modern e-commerce. Amazon estimates that 35% of its sales are driven by product recommendations.

Types of Recommendations

Collaborative filtering

Recommends products based on what similar customers have purchased. “Customers who bought X also bought Y.” Requires a certain volume of transaction data to work effectively.

Content-based filtering

Recommends products with similar characteristics to those the customer has shown interest in. If a customer has viewed black running shoes, other black running shoes or athletic shoes are displayed.

Popularity-based recommendations

Displays the most popular or best-selling products. Simple but effective for new visitors, where there is no personal data to draw upon.

Context-based recommendations

Recommendations based on the current context, such as season, weather, time of day, or the page the customer is on.

Where are recommendations displayed?

• Product pages: “You might also like…” or “Other customers also bought…” below the product description.
• Cart/checkout: “Complete your purchase with…,” perfect for accessories and complementary products.
• Homepage: “Recommended products for you” for returning customers.
• Search results: Supplement search results with recommended products.
• Emails: Personalized product recommendations in newsletters and order confirmations.
• Abandoned cart emails: Show the products the customer left behind, plus related alternatives.

The Impact of Product Recommendations

• Increased average order value (AOV): Customers who click on recommendations typically add more products to their cart.
• Better conversion rate: Relevant recommendations help customers find the right product faster.
• Increased engagement: Customers spend more time in the store and view more products.
• Improved customer experience: Personalization makes the store feel relevant and attentive.

Product Recommendations in Shoporama

Shoporama supports product recommendations through:

• Related products: You can manually link products to each other as related. The relationship is a flat many-to-many association, where each product can have any number of related products, and each relationship can be weighted to control the order.
• Similar products: Typically used to display alternatives from the same category or brand.
• Landing Pages: Dynamic product lists based on categories, labels, and rules, which function as automated recommendations.
• Email recommendations: Shoporama’s newsletter system features dedicated product blocks that display products from selected categories or landing pages.
• Order confirmation recommendations: A dedicated template for product recommendations in post-purchase emails.

Best Practices

• Relevance over quantity: 3–4 relevant recommendations are better than 20 random ones. Quality is more important than quantity.
• A/B test placement: Test whether recommendations perform best below the product description, in a sidebar, or after the “Add to Cart” button.
• Avoid recommending the same product: Do not show the product the customer is already viewing in the recommendations.
• Show images and prices: Product recommendations with images and prices convert significantly better than text-only links.