Search Beats Recommendations, By The Long Tail

We’re all familiar with recommendation systems, due to Amazon’s early efforts as well as current entertainment and e-commerce sites. You visit a particular product page and find related products that might also be of interest to you.

Apparently these recommendations don’t really work as well as search results, in exposing the long tail. According to a recent Wharton School study, current recommendation systems drive less diverse product selection overall. That’s just the opposite effect which so many online retailers are striving for in the first place!

It turns out that simple popularity, an oldie but a goodie, is the biggest driver of recommendations. Says the study, “As long as a product has modest sales, recommendations have the potential to make it a near hit. However, the lower selling products stand little chance at being made famous by the recommender.”

The irony is that recommendation systems may actually expose individuals to more products, with roughly one-third learning about new offerings. However, the perception of diversity may actually be due to search engines themselves, as “consumers have difficulty separating their effects from recommendations.”

Of course, Search is all about the long tail. It’s true that popularity feeds into relevancy as well, based on keywords. While favored URLs may appear first, there are pages and pages of search results available without limitation. There’s no consolidation taking place, as with recommendations.

Still, with the newer collaborative technologies on deck, it’s possible that both Search and Recommendations can further shake out their popularity biases. This way, both can continue to improve their delivery of long tail results.

More: Sign up at the Social Science Research Network and access the complete findings.