The Science Behind Sociocast
Behind the monitor, we at Sociocast have worked tirelessly to solve three interrelated problems: (1) making the online experience as personal as possible, (2) enabling publishers to deliver the most relevant content and (3) helping advertisers reach the right audience in real-time - the people likely to be most receptive.
These problems are difficult to solve, but we're making real progress one success at a time, armed with three core beliefs that lie at the foundation of our science:
- Humans have habits - recurrent patterns in what they do that, when discovered, make it possible to predict future behavior with remarkable accuracy.
- Humans are social - that is, our behavior is influenced by our relationships and social milieu. If properly applied, the behavior of others can be used as a strong proxy for our own tastes and preferences.
- Humans are contextual - meaning that our behavior depends on where we are as much as who we are. Knowing where we are (and whom we're with) sheds a great deal of light on where we'll go next.
These core beliefs have always guided what we do at Sociocast. They allow us to approach our work with clarity and purpose.
Enough of an explanation for you? Bail out here and learn more about our company. To geek out... read on:
What does mobility mean online?
The Sociocast Hypergraph
Social Dynamic = Relevance
Further Research
What does mobility mean online?
Human mobility, an area of research examining the movements of individuals and societies, has focused until now on how people move in the physical world. Data that drive mobility studies are typically gathered from mobile phone records. From these, researchers can discover patterns by seeing where people stop to make calls over time . Studies reveal that the sequential order of an individual's movement has enormous predictive power. It can be used to very accurately forecast where they are likely to be next.
The question, of course, is whether the same principals can apply in the virtual world. Can we, by tracking user movements online, predict with any accuracy where users are likely to go next? The answer is, yes.
Of course, there's a hurdle. While it's easy to define "location" in the real world, it's harder in the virtual world. What, after all, does "location" imply? A website? A topic? A profile page?
Clearly, we need to define "location" to apply the metaphor of mobility online - and this is precisely the problem we solve with a proprietary process we call Context Detection.
For a deep dive on Context Detection, see our blog post Social Mobility (Part I). But suffice it to say that it enables us to discover contexts all across the Web. These contexts help us quickly understand individual consumer interests and intentions, because they represent individual "locations" throughout the virtual world.
The Sociocast Hypergraph
Based on this, Sociocast is able to create a "hypergraph," a kind of map of a social network and relationships between individuals within those networks. We call the map a hypergraph because it's something quite special - a graph that's able to express multiple relationships between individuals (or nodes). These relationships form the contexts that influence individuals, in the same way that, in the physical world, social contexts drive our interests and affinities. In addition, we enable our hypergraph to reveal not just context, but direction - to show the transfer of influence between people in the network.
The Sociocast Hypergraph gives us a rich network representation, one that best enables us to predict how the network will change and evolve in real-time.
For more, see our blog post On Social Mobility (Part II)
Social Dynamics = Relevance
Have you wondered why Collaborative Filtering works, or why you can often count on the "wisdom of the crowds?" Do you find it remarkable that people you've never met can influence the content you seek, and sometimes even provide you with good and relevant product recommendations?
We don't take these questions lightly; they're the focus of our work - of our striving to understand how collective behaviors shape individual preferences, and what strategies will enable us to build systems far superior to collaborative filtering.
Sociocast is just such a system, because it is focused less on what people say than what they actually do. It enables us to find granular audiences and predict how those audiences are likely to change. Specifically, it helps us understand how some people influence others in a given network, and it enables us to proxy some individuals to reveal the interests and preferences of others. The key is our ability to discover the contexts that drive interest and affinity, and to infer social relationships from the billions of user transactions we observe.
This gives us a big advantage over traditional filtering methodologies.
To learn more about this subject, see our blog post Social Mobility (Part III)
Further Research
[1] Barabasi et al., Limits of Predictability in Human Mobility, Science, February 19. 2010
[2] MIT's Reality Mining Project Home (lots of great research here)
