Commentary

Searching For Facebook's Future In Ad Targeting

Facebook has long debated with reporters and bloggers, contending that the social network does not match or track search keywords and search terms to advertisements on the site. Nor does it target ads based on behavior, it says. In recent years I've kept an eye on anything that remotely resembles technology that would allow the site to serve up and match ads based on search terms, but those who support Facebook media relations insist the site's algorithms do not make the match. Well, maybe not today, anyway.

While two patent applications updated this month the U.S. Patent and Trademark Office don't suggest Facebook's technology matches keyword search terms to advertisements, they do describe how Facebook might analyze member messages to identify popular topics and words that contain similar references in advertisements from brands.

facebook/hip-hop

advertisement

advertisement

SEO by the Sea founder Bill Slawski, who first identified the updates of two patent applications, says the patent filings also describe how member-supplied demographic information could correlate to those topics.

The patent application No. 20100164957 filed in December 2008 and updated on July 1, 2010, tells us about how members in public forums often mention and comment on certain topics in discussions. These comments are analyzed to get terms that co-occur with topics. The terms are ranked in frequency and top terms selected based on their rank.

It also describes limitations and short comings of opinion polls, surveys and focus groups before discussing benefits from mining data in messages sent by members of the social network. Slawski thinks the topics could come from advertisers because,  with one filing, a flowchart identifies them as a source.

The patent application No. 20100169327 filed and updated at the same time as the other, describes collecting a list of keywords corresponding to topics and analyzing how members of social networks perceive them during normal discourse with other members.

Facebook isn't interested in common words or phrases that occur in the message along with the keywords provided by advertisers, according to the patent filing. The description explains the importance of the frequency of words or phrases that co-occur when analyzed and weighted based on estimated significance. Only the significant words or phrases, as determined by Facebook and the advertiser, that co-occur within the top terms are identified for analytic purposes.

And unrelated to the patents, Facebook on Friday announced a way to find "interesting pages." Suggested Interests recommends to new members who sign up for updates from "like" publishers with high reader engagement and subscribed-to by people demographically similar to themselves. These new members will see a list of pages that other people in a similar demographic to them also commonly like. This puts the social site at the crossroads between recommendations and search.

So how close has Facebook come to serving up ads based on matching search terms to advertisements? You decide.

Next story loading loading..