What is a recommendation system?

At a glance

• Two main flavors: collaborative filtering (people like you also liked this) and content-based (more items like ones you already enjoyed); most real systems blend both.^[1]
• Big money: recommendations drive about 35% of Amazon’s revenue^[2] and influence roughly 80% of what people watch on Netflix.^[3]
• It runs on data: the more a customer browses, buys, or rates, the sharper the suggestions get.
• The cold-start problem: new customers and brand-new products have no history, so early recommendations are weak until data builds up.^[4]

The two ways it learns

Collaborative filtering finds customers who behaved like you and recommends what they liked but you haven’t seen. Content-based filtering looks at the items themselves and suggests similar ones to what you already chose. Combining them (a hybrid) covers each method’s blind spots and is what most major platforms actually use.^[1]

Why it matters for your business

Good recommendations lift average order value through cross-sells and upsells, keep customers engaged longer, and reduce churn by always showing something relevant.^[2] The catch is the cold-start problem: new shoppers and new products lack history, so you lean on broad popularity or basic profile info until enough behavior accumulates.^[4]

Bottom line

A recommendation system is an automatic salesperson that learns each customer’s taste from their clicks and purchases, then shows them what they’re most likely to buy or watch next.