I very much agree with Fred Wilson and Albert Wenger that we construct various social networks, not just one, to suit various needs. One’s Facebook social graph is often the group of people with whom you are comfortable sharing light-hearted info and photos. One’s Foursquare graph is often more limited: the group of people whom you want to know your physical location. Your Twitter graph is really an interest graph and I think, over time, less about people and more about information sources. Your Venmo graph is even more limited: those people with whom you exchange money and trust. Facebook, I think, could have developed additional functionality sooner to provide impetus to construct these additional graphs (often a subset of your FB graph) directly with them, but I think they missed this. They have been too slow to recognize the importance of physical location, were late in allowing asynchronous following, and are now getting around to payments. None of this is likely to impact their continued massive success, but it opened up opportunities to others.
Missing from this discussion is what I find most interesting about social media: the graph most important to monetization. There are interesting businesses being built on top of the graphs discussed above. However when you really want to build an enormous media business on top of the ridonkulous amount of social data available on the web, you find a new graph emerges: the brand social graph.
At Media6Degrees, the company has pioneered the ability to assemble custom audiences for brands from the connections between their customers and those customers’ friends. For any particular brand, the company actually computes (all anonymously, of course — no personally identifiable info is involved) a brand social graph, finding the friends of a brands’ customers most likely to be interested in actually buying the product. It turns out this is much harder than simply advertising a product to everyone in someone’s social graph. We have multiple connections between us as people, and some are more predictive of mutual brand affinity than others. The brand graph can change by the day, ad campaign, product or offer. This is a hard data problem, but one whose rewards are enormous.
The brand social graph is a big idea. I think, as social media matures, friend targeting via a brand social graph will be a significantly large business model for the largest media companies. Perhaps one qualification: the search targeting model dominated by Google is still likely the most effective form of advertising ever created: the user’s intent is entirely clear, so advertising works really well. But it isn’t great, as we all know, in building brands, finding audiences and creating demand. Brand spending and demand creation is a much larger part of ad spending than demand fulfillment is. I think these needs will largely be served by friend targeting in the future, and one which Facebook is in the position to dominate (but it not yet doing). Google, Yahoo and Microsoft all have enormous amounts of relevant social connection data but do not yet show evidence of using it. Many of Facebook’s announcements last week put it in the pole position of delivering web-wide friend targeting when they deploy an ad network on top of FB Connect one day. And Twitter’s interest graph also gives it a great position in this emerging ecosystem.
I’ve made you blog required reading now after this post! But you know what the 64 billion dollar question is? The elephant in the room? How do you build this stuff without creeping people out?
Building these things is an exercise in software engineering and, to a lesser extent, large-scale hardware operations technology. The algorithms exist, they’re fairly scalable, and hardware is so cheap that the dominant costs become programmer time and electricity.
But it creeps people out when GMail knows enough about AI algorithms to suggest a vendor to you in an email discussion with a colleague about – well – AI algorithms. It creeps people out when they read about credit card numbers showing up in Google search. It creeps people out when US Senators call Facebook on the carpet for their loose privacy practices.
I think you build it by putting people, rather than computers, at the center of the decision loops. Opt-in reduces creep factor.
Hi Ed,
Thanks for your compliments! I agree that general privacy concerns make this challenging, but there are ways to do this, as M6D has done, without using PII at all, so it becomes more of a messaging challenge than an actual one.
David
By “messaging”, do you mean “cooling off overheated privacy advocates?”
Yes, there are ways to do social graph marketing without PII, but there are also ways to figure out from anonymized Netflix data for a contest that someone is gay who has not come out. There are ways to mathematically figure out a non-zero number of peoples’ Social Security numbers from publicly available data. And so on.
There are a lot of smart people in this world. Couple that with the relentless patience of a single modern computer or a botnet, an explosion of self-published content and the innate trust people have in each other most of the time and some unpleasant things can and do happen.
Most of the time, we manage to find a middle ground between utopia and dystopia as we turn our science fiction into reality every day. Sometimes, we don’t, though, and the utopian case is much rarer.
But hey, I’m giving away my Cyborgcamp lecture. http://cyborgcamp.com/
You are forgetting Linkedin.com which is building the biggest business social network. I would dare say that Linkedin.com users will turn out to be more valuable that facebook users