07 November, 2010

Wait until Facebook starts mining our tagged photos...

Many words have been written on the ad revenue that could be generated from Facebooks knowledge of the deep set of factors we provide to them when we sign up for the service. The knowledge of our interests and disinterests as explicitly shared in our information pages and as indirectly shared through our interactions on the site is a major unique boon for Facebook's ability to target algorithms that will have a unique ability to present relevant ad's to users when they are most likely to find the ad content interesting or useful.

This however is only the beginning of ways that Facebook can target the data we are providing them. One potential ad targeting gold mine lies in how we "tag" individuals in our galleries. Facebook could conceivably observe over time the tagging patterns to determine real world associations between people tagged who often share the same photos. It could for example determine if two people are in a relationship over time by watching the tag pattern of photos that mutually contain them. It could also over time predict when relationships start and stop...this is starting to sound very big brother like but it is simply enabled by the fact that we post photos of ourselves, others post photos of us (when we are in their photos) and then others still "tag" those photos. It is true that people can opt out of tagging but for the most part the patterns of our appearance in photos with others over time can be used to determine various types of relationships.

How can this be useful to ad targeting? Well if Facebook can determine through tag proximity analysis (as I'll coin this analysis method) the high likelyhoood that two people are either related or in a relationship they can use that measure to refine their other ad targeting algorithms. This would allow them to for example target ads on the profiles of two people who through their tag associations have shown some affinity...on the chance that the things that one person likes the other (with whom it is **known** physical proximity is shared) likes as well. This algorithm could also, once able to predict that two people are in a relationship can make suggestions when mutually important dates to these individuals arise...for example, it can suggest gifts for individuals when their birthdays come up with a strong likelyhood that those individuals are in a relationship. All of this would be done in the back ground and the user would just notice that Facebook's ads are unusually accurate and it would come out of using the tag data that is being provided by the users and their friends.

Recently Peter Thiel former executive at PayPal made the prediction that Facebook was severely undervalued and I think he may be right, mining tagged photos for behavioral information that can be used to target ads is only one of the many tools that Facebook is enabling and that people are freely contributing unique data about their social interactions to the service through...it will be interesting to see what will be as time goes by.