The design and concept of the previous version of this website was well received, and when I made the decision to replace it with something more functional I knew I wanted to continue playing with the idea of somehow representing how we interact with online services. The result was Atsumeru, (the Japanese verb to collect), a web app that allows anyone to collect and represent their interactions with popular social apps.
The initial idea was to display information in much the same was as on my personal site, using metrics such as activity per day, number of words written, most listened-to song etc. and presenting each user with a breakdown of how they had used the service. I refined and extended the code I wrote, for example to account for the difference in usage between services, I implemented a scaling factor. Applied to each service, it weights activity that take longer to ‘create’, such as blog posts or photographs, against activity on services like Twitter and Last.fm.
I also expanded on the idea of breaking down data by timeframe so activity could be seen not only for the lifetime of a user's account but also over arbitrary ‘periods’ – the time I've spent employed at each of my former companies for example. Users can flip through these and see how my usage changed over time, what music I was listening to and so on. I also like the idea of expanding this further by allowing users to add a location to a period, so activity can be geocoded to some degree, much like the map panel on beseku.co.uk.
After learning and messing around in Rails for a few weeks, I came up with a first draft … and it doesn't work. After asking a few friends to test it and add their data to my own, it's obvious that not only is the data indecipherable, most of it is pretty uninteresting. I took a huge misstep in trying to compare services against each other, as the interactions are so very different. Writing a blog post is far removed from tracking what you listen to. The use of a scaling factor made the graphs look prettier and more even, but unless you know how it works it makes scale so much harder to understand.
What I did find though, is that the idea of timeframes to allow users to divide their interactions into memorable periods was interesting. Just seeing how I used Twitter more when it launched and less so now, or that I listened to way more Broken Social Scene in 2009, made me want to be able to explore further. Building this first prototype has allowed me to identify what parts of the product work and what failed, and allowed me to adjust the direction that I head in while still having lots of data to play with and the basic structure app in place. I'm excited to see where it is heading.