About Amazon recommendations

Amazon.com started as online bookseller, and has grown into one of the most popular online stores, carrying a vast array of both new and used merchandise.

Description of user contribution system

Amazon presents shoppers with recommendations of other products to buy which are automatically generated based on the behaviors of other users.  In this way Amazon uses a user contribution system to “merchandise” the site (ie select which products to show) for each shopper uniquely. 

Take a look at this page marking up all the ways Amazon employs UCS, or visit Amazon.com and look for these things on the item details page:

• Customers who bought this item also bought
• Customers who viewed this item also viewed
• Customers who bought items in your recent history also bought

The system is passive as Amazon customers do nothing: their buying behaviors are automatically collected by Amazon.  Amazon presents recommendations at several points in the shopping experience:  on the home page, on the “item detail page” that describes each item Amazon sells, on the page confirming successful checkout.

The recommendation is specific to the product the shopper is viewing or has bought.

Benefits for the users

This system helps the shopper find relevant items to buy in addition to, or instead of the item the shopper is viewing or has just bought. (The contributors need no benefit since they do nothing and may not even know the system is using their purchase patterns to the benefit of others.)

Benefits for the creators of the system

The system helps Amazon generate higher sales in several ways by suggesting additional purchases relevant to the shoppers’ interests as evidenced by their shopping and buying behaviors.

History of the idea

Started as Instant Recommendations by Greg Linden at Amazon. Greg summarizes the beginning on his blog.

Scale

While it’s not known what percentage of Amazon sales are credited to Amazon recommendations, Amazon has several million “item detail pages” most of which feature one or more types of recommendations. 

More resources

• The “secret” behind Amazon recommendations

• Amazon.com Admits Concocting Some Recommendations

• Amazon.com Recommendations: Item-to-Item Collaborative Filtering, a technical article in IEEE Internet Computing by Greg Linden ($19 to access fulltext)

See also on this wiki 

• Buying behavior