What percentage of eCommerce sales are driven by recommendations?

An estimated 35% of Amazon’s revenue can be attributed to its recommendation engine alone. How can a recommendation engine really gauge what someone is more or less likely to buy?

The important thing to keep in mind is that no single recommendation strategy is equally effective for all site visitors. Some people visit a site frequently and have high interaction levels, allowing a recommendation engine to accumulate a wealth of data about their buying and browsing habits (as Zoran has pointed out). Others may be first-time visitors who may have arrived via a search query or a social media promotion, which affords minimal visibility into their tendencies and preferences.

To overcome the challenge of engaging different types of customers, the best recommendation engines first assess the level of valuable information about each visitor and automatically deploy the most appropriate strategy based on behavioral data, context in the purchase funnel, or general popularity signals on the site.

Here’s an example of different recommendations strategies as they can be applied to engage different types of consumers:

Personalized Product Recommendations

To truly present products that each visitor is most likely to buy, recommendation engines leverage two powerful strategies: collaborative filtering and affinity-based recommendations:

Affinity-based recommendations: Recommending products based on a weighted score of each consumer’s onsite interactions (viewed, added-to-cart, purchased)

Collaborative filtering: Filling in recommendation slots with products that similar consumers have bought or engaged with.