April 1, 2008

Twitter id's and number of Twitter users

A comment on the previous post (1 million Twitter users) stated that "there are already a lot more than 12 million accounts on Twitter". The proof of this statement is given by information grabbed from the Twitterholic site.

Each Twitter account has its own Twitter id. You can easily find your own Twitter id by looking at the URL for the RSS feed for your Twitter account. In the early days of Twitter, Twitter id's were sequentially attributed. This made it easy to track the number of total Twitter accounts created (not necessary the number of active Twitter accounts). Jason Kottke published almost a year ago an article based on such stats.

Because of the growing popularity of Twitter, the Twitter developers decided to change the policy of generating Twitter id's by the end of November 2006. From that moment on there were gaps between the id's actually used. Assuming that the Twitter id of a new created Twitter account is equal the total number of Twitter accounts does not seem correct to me. The exact number of Twitter accounts created until now (and the number of active Twitter accounts) is only known by the people at Twitter Inc. And they haven't disclosed it until now ...

By the way, currently it is not possible to register Twitter names containing spaces. In the early days it was possible to register such Twitter names. Some of these names are still in use, e.g. twitter.com/there you go.

1 comment:

Kevin Makice said...

I took some samples of the public timeline about a year ago and examined the unique user IDs. Although the pattern of ID assignment changes, in the early days at least Twitter would alternate between from sequential to non-sequential, with the latter being in increments of 10.

So the first 13,753 were sequential, followed by a run up to 754363 where they likely incremented by 10s. At 754439 they again went to sequential until 824222, and at 824331 through 5281761 again incremented by 10s (last digit ending in 1). In my dataset from that period - which is not inclusive of all accounts - showed increments by 10s with the last digit ending in 2.

I strongly suspect that a respectable estimate would be to simply divide the maximum ID by 10.