Facebook has filed a patent that would make it easier to target whole families with ads by analyzing the photos they post. The application, filed on May 10th and published today, covers an algorithm that would identify elements of photographs — like faces or other details— and cross-reference them with other data to build a profile of an entire household. If Facebook chose to implement the system, it would supplement a family targeting program that launched last year.
Facebook can already analyze lots of information to tell who’s in the same household. According to Marketing Land, it checks the relationships people list on their profiles, whether people list the same last names or locations, and shared life events or event check-ins, among other things.
The system described in the new patent would involve an even more sophisticated level of data mining. The proposed model would cross-reference details from photographs with tags, descriptions, the poster’s IP address, the list of Facebook users using that same address, and potentially other details. From this, Facebook could deduce how many people were in the household, alongside various demographic information. It doesn’t specify what that might be, although advertisers often look for data on gender, age, and socioeconomic status.
In one example, Facebook describes a male user who posts multiple pictures of two female subjects who appear repeatedly in either pictures he posts or pictures friends tag him in. One of them is a picture of a single young girl, with the description “my angel.” The system could deduce that there are three people in his household, one male and two female, and that the female members are probably his wife and daughter.
That may or may not be accurate, but it would let advertisers target that user using those demographic characteristics — even if he didn’t explicitly list relationships with his wife and daughter, or if the wife and daughter aren’t on Facebook at all.
The application’s claims are broad, so this hypothetical system could be either extremely specific — like cross-referencing faces — or extremely comprehensive, like analyzing photos for brands or other signifiers. (A previous application covered turning on phone microphones to detect what TV shows users were watching.) Facebook doesn’t go into what “demographic composition” covers. The description touts a way to hyper-target users based on household size, household member characteristics, shared interests, which household members use which electronic devices, and whether those users are Facebook members.
Facebook has put a lot of emphasis on how its platforms can bring families together. It recently launched its Portal home video chat device, which is aimed particularly at helping people connect with children or grandchildren. Incidentally, the Portal includes advanced face-tracking features for its “smart camera.” Facebook has said it might use Portal call and app data to target ads on other platforms, and the idea of analyzing images the same way might be tempting.
This patent covers photos that are explicitly posted by users, not video or content from private messages, and companies frequently file patents they never act on. So while Facebook probably could use your Thanksgiving family photos to pitch more attractive Black Friday deals next week, this patent doesn’t tell us whether it actually will.