Counting Beans on Twitter

In my post “So Tweet This Maybe?” I tell the story of how one little tweet made a noticeable difference in the size of the audience alerted to a particular story online. Here’s a bit more detail on how I estimated the reach of this tweet:

#AAASmtg A fun, quick read by @ on placebo effects of wind farms. Yes wind farms. Go read it:

If you click on the date/time stamp in that embedded tweet, it will take you to Twitter, where you’ll see that (at the time of this posting, anyway) 16 people retweeted it directly, and three people retweeted using other methods or with added commentary. Here is what I saw when I looked at it on my iPad (you’re always getting slightly different views, depending on your platform):



You can click through to each of these user’s profiles to see how many people are following each of them. When I ran the numbers on February 19, 2013, this was the follower count for each person who retweeted me:

Ed Yong 30,300*
Joe Hansen 7,971
Matthew R. Francis 2,890
Colin Beverage 1,663
ScienceWriter 1,534
Tanya Noel 1,088
Cristy Gelling 641
Susannah Locke 495
James 439
Y.G. Mitchell 416
Carol Morton 334
Cameron Walker 267
Tim Cross 184
AMSTAT Philly 147
Warren J Code 87
Allison M. Hydzik 61
Roxanne Palmer 54
Liz Rauer 41
Jessica Higgins 14

All together, plus my own 3,903 followers at the time = 52,529 impressions! Impressions represent the total possible reach of a story. Given that people often miss tweets while they’re offline, and that a single individual might have gotten many of the subsequent retweets, it is important not to think of this as if it were 50,000 people.

But, it’s also important to note that these 50,000 impressions are only indicative of first-degree relationships. I didn’t go through exhaustively and count the secondary audience created when Ed Yong’s RT was RTed by 14 more people, and Joe’s was RTed by 4 more… so the ‘true’ number of people reached is even larger than those counted by the 50,000 first-degree impressions. Think of it like ripples in a pond. I just looked at the first wave.

My favorite part of this story is that the conditions could hardly have been better to run this little experiment: There were four days between when the story was posted and when I tweeted about it. I am the only person Erik asked to tweet about it. The twitter date and timestamps, retweet links, and google analytics combined make a pretty compelling case for causation.

Am I wrong? Missing a key element in the analysis? I would love to hear your feedback!