Posts Tagged 'Loomia'

Amazon’s Page Recommender Widget Moves Into Content, Challenges Start-ups

Amazon has launched a new Page Recommender widget that affiliate bloggers (and presumably other publishers) can put on their sites.

The widget makes personalized recommendations of other pages within a Web site to users, apparently based on their individual preferences. Whether the recommendations are based on 1) just their browsing history within that particular site, 2) also their browsing on other affiliate sites, 3) their product preferences within Amazon.com proper or 4) all of the above, is unclear.

In addition to the new content recommendation, the widget also recommends products. The product recommendations are also personalized based on individual preferences. From Amazon, via O’Reilly Radar:

The products that are recommended are based on the interests of each individual visitor as well as items that convert well for your website. The products that you see may be different from the ones that are displayed to others based on their individual behavior.

This move into content recommendations is not too suprising, since Amazon’s product recommendation engine on its site has been one of the best examples of recommendation engines. That is, those parts of the site that say “readers who were interested in this book were also interested in this.” In theory, using that recommendation engine for Web pages instead should not be too big of a change.

Because this is a new form of affiliate revenue, to some extent this competes with Google (Google and Amazon increasingly compete in other areas such as cloud computing), but more directly it takes on a number of start-ups that are providing similar or related services. These companies, such as Aggregate Knowledge, Inform and Loomia, and Proximic, and to a lesser extent Sphere, provide recommendations for related content and/or products based on the history of traffic and click-throughs in their networks. Some of them can do nifty things like recommend a product in a completely different area from an article that someone has viewed. For example, on a page about gardening, the service could recommend a specific type of bicycle.

However, as far as I know most of these start-ups are not basing their recommendations on personalization as Amazon seems to be, but rather the aggregated knowledge of their networks. There is an interesting strategy difference in presenting the most popular versus the most individually customized recommendations (see my previous post).

The Amazon widget also falls into the larger trend of ads and content merging together (Google’s AdSense is the most obvious example of this. When you search for shoes, the distinction between the unsponsored search results and the sponsored links has become less important for users.)

This could the beginning of a broader push into content recommendations for Amazon. However, I’m not sure this is a big challenge to Google, I think it’s a much bigger concern for start-ups.

Social News: Popularity Vs. Personalization

ReadWriteWeb has an interesting post on “collaborative filtering” on social news sites. The writer makes the point that there are two types of social news recommendation models: one that provides you with the most popular stories from all members of the site, and another that provides you with stories that are personalized to you based on your past reading habits and those with similar interests as you.

There are different camps here: Digg based on the former model of popularity and others such as Reddit, Stumbleupon and Searchles that recommend stories based on various forms of personalization. (There are also services such as Aggregate Knowledge and Loomia, which use some aspects of social recommendations to recommend related articles or products.) However, the two methods, popularity and personlization, are not mutually exclusive. Digg is about to implement a recommendation feature. And many sites, from big media sites to social media sites, could provide both a most popular section as well as personalized recommendations.