Recommendation systems: how “you may also like” sells more in your online store

What is a recommendation system?

It is the part of a store or an app that decides what to show to whom. The “Customers also bought”, “You may also like” and “Complete the look” blocks you see on large online stores are recommendations. Instead of showing every visitor the same products, the store adapts to each person.

Why it matters

The biggest platforms are built on it. In 2013 McKinsey estimated that about 35% of what consumers buy on Amazon comes from product recommendations, and in 2015 Netflix’s product team wrote that about 80% of the hours streamed on Netflix are influenced by its recommendations. You don’t need their size to benefit: the same idea works for a store with a few hundred products.

Before and after: two visitors, one store

Imagine a sports store. Without recommendations, every visitor sees the same home page and the same bestsellers. With them:

• Visitor 1 looks at trail running shoes. The store suggests trail socks, a hydration vest and the shoes other trail runners bought.
• Visitor 2 looks at yoga mats. The store suggests yoga blocks, a mat bag and leggings, not running shoes.
• In the cart, each one sees one small, relevant add-on instead of a random promotion.

Same catalog, two different stores: each visitor finds what interests them faster, and is more likely to add a second item.

How it works, in simple terms

• Similar products: products are compared by category, brand, color, price or description. Look at a blue linen shirt and you see other linen shirts.
• Bought together: the system learns from orders. If many customers who bought a camera also bought a memory card, the card is suggested with the camera.
• Personal history: products are ranked using what this visitor viewed, searched and bought before.
• Popular and trending: for new visitors with no history, the best sellers of the moment, ideally per category.

Good systems mix these methods and fall back on popular products when they know little about a visitor. This is called the “cold start” problem, and every store has it at the beginning.

Where to show recommendations

• Product page: similar products and items frequently bought together.
• Cart: one small add-on that completes the order.
• Home page: recently viewed products and picks based on past visits.
• Emails: products related to a past purchase or an abandoned cart.
• Search: results ranked by what the visitor is most likely to choose.

What you need to start

• A clean product catalog: categories, attributes, prices and stock.
• Your order history and, with the visitor’s consent, browsing data.
• Business rules: never suggest out-of-stock items, respect your margins, avoid showing the same product twice.
• A way to measure: clicks on recommendations, average order value and conversion rate, ideally compared with an A/B test.

Privacy matters. In Europe, personalized tracking needs the visitor’s consent under the GDPR and the cookie rules, so recommendations should still work without it, using only the context of the page being viewed.

Even small stores can start

Start simply with “similar products” and “frequently bought together” calculated from your own orders, then add personalization when you have more data. At Acmonic we build recommendation systems into new online stores and connect them to existing ones, with clear reports so you can see what they bring.

A good recommendation works like a helpful salesperson: it shows the right product at the right moment, and never gets in the way.