A few days ago I got curious about the breakdown of Twitter update sources. I searched around the Internet first and found a set of results from ReadWriteWeb. The results seemed to be somewhat limited and were a one-time statistic taken over a given timeframe and analyzed later.

This wasn’t what I was looking for. I wanted a real-time set of results where I could check on daily trends and compare one day to another in terms of which services see the most use. I also wanted to be able to visualize the data and see the percentage breakdown of those services. In short, I wanted something to give me stats that are always up to date.

If you’ve been following this site at all, you’ll probably realize this is the part where I debut some new thing I’ve been working on. This post is no different. Let me introduce to you….. Tracker.

Tracker works by sampling the public timeline every 60 seconds. Twitter limits the number of response you can get per minute to 20 (and caches that to really enforce it). So, every minute, Tracker goes out and grabs those 20 tweets and figures out what service was used to send the tweet. Tracker stores all that info in my database, broken down by day for trending purposes.

The front-end of Tracker consists of a single interface where the user can view a set of stats for a specified timeframe. The user can view the stats in ranking order by number of tweets sent, and also view the percentage share that service holds. The percentages are divided into two categories: overall share, and non-web share. I chose to create non-web share because the web interface accounts for over 40% of all tweets sent. This way, users can tell what third-party clients are being used and see percentages of just those.

I’m going to let Tracker run for a few weeks to gather enough data to do weekly trending, at which time I’ll loose the URL for it. I’ll apprehensively let the URL slip, just for kicks. Please keep in mind it’s a work in progress, but I’m confident in its accuracy. You can hit Tracker at http://mark.bockenstedt.net/twitter/. Tracker started running on August 6th, 2008 at 10:30 AM and has collected over 50,000 tweets. Over one month’s time, it will sample nearly 864,000 tweets. I can’t find any solid numbers about Twitter’s actual volume, but I estimate this to be around 10% of all the tweets that are posted.