How Real People Use Twitter

I am not a typical Twitter user. I follow over 3,000 people, and I have over 4,500 followers. I say this not to brag (you may have more, certainly), but to note that I am an outlier. Indeed, according to the most recent data I've seen (Twitter Statistics For 2010, produced by Sysomos) 98% of Twitter users have under 1000 followers, and only about a tenth of a percent have more than 10,000 followers. By those standards, I'm kind of a freak. Twitter statistics followers table

The ease with which we can add "friends" followers (with or without their consent) lulls us into thinking that attention on Twitter is a fungible commodity - that there is little cognitive "cost" to the Twitter user to follow just one more person, brand or "celebrity." But 96% of Twitter users follow fewer than 500 people - probably not much different to their Facebook "friending" habits.

Twitter statistics Friends table

That's why mathematical exercises like this one, which compares persons with under 1,000 followers to those with over 1,000 followers, are dangerously deceptive. People with over 1,000 followers are statistical outliers. If we attempt to derive "best practices" by looking at the habits of these outliers (in other words, like the author of the post I cited, if we treat statistical outliers like "goals" instead of anomalies), we run the grave risk of falling for what statisticians call the "Texas Sharpshooter Fallacy."

This term stems from a joke, really - a Texan blasts away at the side of a barn, finds where the highest density of shots are clustered, and then paints a target on them. When we look at the habits of the "Twitterati," that one-half of one percent of users with over 20,000 followers, we unwittingly engage in the same deception. Sure, we can look at their habits - when they tweet, how often they tweet, how often they reply and how often they share links - but when we squint really hard at these graphs, and backtest our way into perceived "patterns," we leave out a pantload of confounding variables. In other words, we can optimize tweeting patterns all we want, but if we want lots of Twitter followers, we'd probably be better off learning to dunk, sing or be President.

Instead, I would challenge us all to look at where the "magic middle" lie, the 96% of Twitter users who follow under 500 persons, and how they choose their "followees." Our EVP at Edison, Joe Lenski, is a political junkie and our expert on all things related to polling and the mastermind behind our efforts behind the National Election Exit Polls. He derives immense benefit from Twitter - really enjoys it - and he follows "just" 120 people. His list is carefully curated and almost without exception targeted like a laser on politicos, pollsters and fellow election researchers.

Getting on Joe's follower list is a bit harder than getting on to mine, and the effectiveness with which Joe uses Twitter has me regretting my profligate ways. Understanding Joe, and all the Joes out there (they are legion) is the key to truly understanding what Twitter can become. This, by the way, is my secret hope for folks like Klout and PeerIndex - if they can extricate themselves from the confounding variables of follower counts and retweets, and continue to beef up their research into topics, authority and true engagement, they'll find the Joes of the world.

For my part, I'm pruning down my follower list to something a little more reasonable. I'm taking a cue from Mark Schaefer and taking responsibility for the quality of my Twitter stream. It's not you, it's me. Chasing after those with 10,000 or more followers and emulating the received wisdom of their "habits" may or may not be a goal for you, but never forget that these Twitter celebrities are outliers, and not necessarily goals. You cannot derive the reasons for their celebrity from mining unstructured data alone, or you're no better than our vainglorious Texan friend. Paint your own target first - then shoot.