Hackathon Project: Twitter Client Use at Chirp Day 1

The daytime part of #chirp was a lot of fun, but the overnight hackathon is another thing altogether. Not only do we get to play with newly-released APIs — user streams, I’m looking at you — but we also get introductions to some pretty bad-ass new libraries like the @ Anywhere JavaScript API from the developers who wrote them.

Oh. And then there’s the rate limit bump to 20,000 requests per hour. That’s kinda neat too.

As a quick first project for the evening, I decided to follow up on a suggestion from @_stritti_ to do some quick analysis on which Twitter clients were popular at the daytime part of the conference.

I wrote a quick Ruby script using grackle to go out and grab the most recent (up to) 2000 tweets from each of @boctor’s Chirp attendee lists and aggregate the source attributes from each to build a chart depicting the popular clients:

So it turns out that the website is by far the most popular client, accounting for 466 of the 2356 tweets analyzed. Interesting. Tweetie, TweetDeck, the generic “API” source, and Seesmic came next with 356, 325, 214, and 98 tweets, respectively. The top 5 clients account for more than 60% of all tweets. So Twitter clients are not a winner-take-all kind of market, but there isn’t room for a whole lot of big players, either.

It should be noted that this is probably not a representative sample of normal behavior of the Chirp attendees. Since the WiFi sucked during the day, there’s probably a lot more mobile tweets in the sample than there ought to be. Regardless, though, it is a good analysis of which Twitter clients were popular at day 1 of Chirp, so I’ll take it.

The code and a spreadsheet containing the results are attached to this post. To run the script, you’ll have to enter the username and password for your Twitter account in the code itself. (I’m too lazy to give them actual command-line parameters.) Also, be warned that running the script will take 30 API calls, and unless you’re blessed (like we Chirp attendees are) you only have 150 per hour. So use sparingly unless you’re terribly curious.

EDIT: Thanks to @grandcentralvc for pointing out that clients are counted by the number of tweets they served, not the number of users. That’s a different data set, which I can build if anyone wants to see it.