Approaching Twitter Demographics

// March 30th, 2011 // Cool Stuff, Life, Social Media Marketing, Technology

Recently I was introduced to InMaps from Linked In. What it does is analyze your Linked In contacts and their contacts to show how your professional contacts are connected. It then colour codes them based on common links. Finally users who are more connected get bigger dots than others. See mine below:

InMap of Dmitri Dawkins

This is pretty cool because now you can see how your contacts are connected. But how else can data visualisation help us?

Location:

I have been trying to track keywords and events based on location via social media tools for a long time. The issue with social sites like Twitter is that people don’t set their location correctly. For instance if a Twitter user enters Jamaica as their location then it assumes Jamaica, New York, USA, not the country. Also people use other terms like jamdown, jamrock, or just cities like Kingston or Montego Bay. When trying to find tweets from a specific area even if the person matches the keywords the location may be off.

So I first considered creating a list of Jamaican twitter users, however how will this be updated (migration etc.)? Then I thought of tracking interactions and then flagging users as Jamaican, but once again that would require alot of human interaction. Ideally I want something automated and dynamic. InMaps gave me a potential answer. Track interactions between users and then group them into subsets just like InMaps. Couple that with a metrics system, where a user with properly filled out Location or listed under multiple twitter lists with specific keywords or Klout has a higher metric than others. When users are associated with other users with higher metrics it helps to create focal points for the network.

Online communities are associated by content, relation (friends, family, groups, alumni) or location. By establishing the link between individuals online we can create a dynamic social mapping tool more accurate than what currently exists, even more accurate than what the user them-self knows. This would allow analysis of keywords by social-location-grouping or possibly by profession, education, and more.

What are your views?

Post to Twitter

2 Responses to “Approaching Twitter Demographics”

  1. Dan Holowack says:

    Great post Dmitri!
    I hadn’t thought of Twitter demographics in this way before – but it makes sense. I often @reply with close friends and family in my geographical area. It may be possible to use a weighted system (and neural network approach – I’m learning lots about machine intelligence these days) to create clusters of users.
    We have some heavy math and stats guys on the TwitSprout team and I’m really interested in digging into some powerful social relationship mapping. Lots will happen this summer.

    Thanks for this post – and the quick cc on Twitter :)
    Cheers!
    -dan

  2. Very relevant video where 12 billion phone calls are crunched to extract clusters based on social interaction.
    http://spatialanalysis.co.uk/2010/12/10/creating-regions-from-social-relationships/

Leave a Reply