Last Updated on 14 April 2010 by gerry
Following my experiment with sampling the public Twitter timeline this morning, I ran another sample, this time for much longer and at a time of day, when many American users would be active to get a slightly more accurate breakdown of the top twenty Twitter clients.
In comparison to the earlier results, there has been a bit of jostling in the mid-to-bottom table positions, but the top three positions remain the same. The percentage of tweets posted “via web” is greater at 37%.
| RANK | SOURCE | % |
|---|---|---|
| 1 | web | 37.3% |
| 2 | UberTwitter | 9.5% |
| 3 | TweetDeck | 7.4% |
| 4 | Echofon | 4.1% |
| 5 | API | 4.0% |
| 6 | twitterfeed | 3.4% |
| 7 | Twitter for BlackBerry® | 1.9% |
| 8 | movatwitter | 1.8% |
| 9 | txt | 1.7% |
| 10 | Tween | 1.6% |
| 11 | HootSuite | 1.6% |
| 12 | TwitBird iPhone | 1.5% |
| 13 | Tweetie | 1.4% |
| 14 | mobile web | 1.2% |
| 15 | Keitai Web | 1.2% |
| 16 | Twitterrific | 1.2% |
| 17 | mobile web | 1.0% |
| 18 | twidroid | 1.0% |
| 19 | Snaptu | 0.9% |
| 20 | Seesmic | 0.8% |
Based on a sample of 119357 tweets acquired on 14 April 2010 just before 16:30 UK time using the statuses/sample Twitter API method and the Phirehose PHP class.
Update: I’ve just discovered that a daily breakdown of twitter clients is presented by @twitstat. Notably their “via web” percentage is much reduced compared to what I measured and UberTwitter ranks in a much more lowly position. Their method is different (better?) in that they only sample unique user/client combinations – I might try that another day to see if my results really are skewed so significantly.