From the social graph to the place graph

Posted on February 6, 2012


I was at the #monkigras conference last week (which was excellent, btw, and writing it with the hashtag is a nervous tick I can’t shake) and promised myself I’d blog about some of the great things I heard.

Matt Biddulph used to be at Nokia, because in 2009 the handsets giant bought his start-up Dopplr, a social networking site for people travelling around the world. I followed Dopplr a fair bit in the couple of years before that because I was elsewhere trying to crack a problem for people who travel, in my case simply how they get online (a problem that has since been solved – to oversimplify – by the evolution of smart-phones and location-based services). My baby was something called, which I started while at CNET Networks.*

Matt got on to discussing some decades-old thinking in social networks – not provoked by  social networks as we think of them now but the type anthropologists and social scientists have long studied – and something he has come to call ‘the place graph’.

The idea is that just as a Facebook or LinkedIn can predict, for example, people you might want to know based on the relationships and interactions you have exhibited, so too a service might be able to make suggestions based on place.

At Nokia, Matt used a Hadoop cluster to process millions of navigation requests and build up what he calls the “real map of the world”.

The map was fascinating but in part merely tells you about a location based on requests, logged by the Nokia servers, from people travelling to or from it by car.

But the clever part comes when you can start to ask ‘Where is the Shoreditch of Madrid?’ or ‘Where is the Greenwich  Village of Tokyo?’ (To depart from Matt’s script just a little.)

Matt describes it way better than me and you can see more here: .

Network relationships don’t just have to be about people. There are plenty of budding local and hyperlocal businesses out there now that would love to plan based on this kind of geo-like-mindedness.

*It’s still just about there if you look but not supported and far from in its intended state. I doubt it’s long for this world.