Lets face it: no matter how much we protest that Twitter is not a numbers game, we can’t help comparing follower counts with other similar accounts. And we all get a little fizz of excitement when we pass a milestone in followers, or notice a spike in the number of people following us.
But if twitter really isn’t a numbers game for you, it can be interesting to look at how your follower count has changed over time. This post compares at three different ways of estimating your growth in followers.
1) Twitter’s own analytics
Most twitter accounts I’ve tried can access at least some of their Twitter analytics: https://analytics.twitter.com/accounts, even if they haven’t had a notification that it is available to them. I get most of the same analytics on my personal account as I do on the work account – the only difference is I don’t have a link from the “settings and help” cog in the top right hand corner of the twitter home page on my own account.
The second tab along shows some information about how many followers an account has:
This graph is not the number of followers you had each day – if you have a mass exodus of followers, the graph will still only show growth, as it only shows when your current followers started following you. For more, see Twitter’s help.
Although you can hover points on the line and see how many followers started following you on that day, there is no way to export that data (annoyingly). What you can do, however, is copy them all into a spreadsheet then plot them yourself…
2) Accession analysis
Using the website ScraperWiki, you can download a list of any twitter users followers. Included in that information is the date they joined twitter, and the list is in the order in which they followed your target account (most recent followers first). Using that information, and the fact it is impossible to follow a twitter account with an account that doesn’t exist yet you can plot an “accession analysis” (basically their follower position against the date they joined twitter). For more on this technique see the oUseful blog.
Here is that graph for the account I referred to above:
The second jump in followers is the jump shown on the 25 November (actually mid December 2012) on the first graph. This is still a function of the accounts that are following the account at the point of measurement – it still won’t show a mass exodus of followers.
3) Keeping a daily record of follower counts…
We use a website called TweetReach to collect (amongst other things) every tweet we’ve sent. One of the cool side-effects of TweetReach is that it also records how many followers we had at that time (as a measure of “reach” of the tweet). By plotting that data we can see exactly how many followers we had at that point in time:
The green line shows the accurate “followers at the time of the first tweet on each date”. The blue dots are the plots of “accession vs twitter join date” and the crosses show the plot of “started following us date” from twitter’s own analytics (copied out and plotted by hand).
So which technique to use? Conclusions…
In this case we are lucky that we haven’t had a massive drop off in followers, so the plots resemble each other very closely. In actual fact the first two measures only look at the followers that follow us currently – they are not a true measure of how many followers we had on each day. So long as the number of daily new followers is greater than the number of daily unfollows, the graph will remain broadly accurate. As soon as you start to piss off lots of followers, that’s when the graph stops being accurate.
The second two analyses don’t require access to the password for the account you are checking – so these could be useful for comparing your followers with a competitor’s account, or to spot if a twitter account might be buying followers – a leap in followers doesn’t just happen spontaneously, and if there is no obvious cause, that suggests followers being bought. In this example the first spike corresponds to the London disorder in August 2011, and the second spike was a series of humorous tweets about icy roads.
Although the accession analysis is a little less accurate on the number of followers (underestimating it a touch in this case), it goes back further than twitter’s own analytics (which only goes back a year and a half) and the Tweetreach data (which goes back only as far as you have the data for!).
The third approach only works if you have access to the data – and if you haven’t been collecting the data, you might struggle to get it. TweetReach do offer a historical search tool (giving access to every tweet ever written) but there is a cost and I don’t know if it includes the number of followers at the time the tweet was sent.
The ultimate answer is “it depends”. Different scenarios will require different approaches – and there are pros and cons to each!