Monday, December 11, 2006

35% of sales from recommendations

In an article about startup Aggregate Knowledge, Matt Marshall writes, "Amazon says 35 percent of product sales result from recommendations."

6 comments:

Anonymous said...

The factual errors in that article are too numerous to count.

Greg Linden said...

Are they?

The 35% number likely is accurate. Personalization was responsible for well more than 20% of sales when I left Amazon in 2002. I would be surprised if it was not even higher now.

However, the article does claim that Amazon only rebuilds the databases used for recommendations every month. My understanding is that they are rebuilt multiple times per week (and the recommendations any individual shopper sees change immediately on new ratings, clicks, or purchases), so I think that part is unaccurate.

Are there other inaccuracies?

Eran Sandler said...

Haven't you considered generalizing and out sourcing some of the technology behind Findory to do just that?

Gabe said...

I'm curious: at what rate are recommended items purchased?

Specifically, if an Amazon user views a page with recommendations, what are the odds that a given recommended item is purchased by that user? 1%? 0.1%?

AjiNIMC said...

I think the web owners should try to provide platforms encouraging recommendation (including some stat capturing).

IMO recommendation is the best unit of measurement for quality.

Greg Linden said...

Gabe, sorry, I do not have those numbers.

Eran, that is a good idea. I have been thinking about whether Findory should attempt to launch a recommendation web service. The idea would be that, given some content (news, weblog posts, podcasts, etc.), Findory would make recommendations for other interesting content (news articles, documents, advertisements, etc.).