Science May Have Cracked the Code to the Perfect Tweet

The most popular tweet of all time needs no introduction. 

If only Bradley's arm was longer. Best photo ever. #oscars pic.twitter.com/C9U5NOtGap

— Ellen DeGeneres (@TheEllenShow) March 3, 2014

Retweeted more than 3.4 million times, Ellen DeGeneres' tweet from the 2014 Oscars made social media history early this year. The celebrity studded selfie trounced the former No. 1 post, President Barack Obama's tweet about his re-election:

Four more years. pic.twitter.com/bAJE6Vom

— Barack Obama (@BarackObama) November 7, 2012

 

One could easily attribute these tweets' success to DeGeneres and Obama's online and offline popularity. But recent research conducted by Cornell University and Google wanted to find out if there were certain content characteristics that made a tweet more shareable, removed from "characteristics of the originating author."

"In other words," the paper's authors write, "we investigate whether a different choice of words affects message propagation, controlling for user and topic: Would user BarackObama have gotten significantly more (or fewer) retweets if he had used some alternate wording to announce his re-election?"

To do this, the researchers looked at tweets that illustrated a common practice on social media: the same user posting about the same link in two differently worded tweets. By controlling for author, topic and a number of other factors, researchers whittled 1.77 million tweets down to a few hundred thousand Twitter pairs—same user, same topic, different words.

After a number of linguistic tests and filters, the researchers say they have a pretty good handle on what features are more likely to get a tweet retweeted. With the algorithm they created from their data, researchers were 66 percent accurate in figuring out which tweet was more popular, compared to 61 percent for the average human. (You can try your hand at guessing which tweet is more popular with their online quiz.)

Some of the characteristics of those tweets that were more likely to get retweeted:

• They asked to be retweeted. A solid piece of advice works on social media as well: If you want something, ask for it! Tweets where users made "explicit requests for sharing" were shared more than the tweets that didn't. Within that, some retweet signifiers were stronger than others. "RT," "retweet," "spread" and "please" were more effective than "pls" and "plz" for getting followers to retweet. 
• They were informative. "Messages that are more informative have increased social exchange value, and so may be more worth propagating," according to the research. Lengthier tweets were deemed more informative, but, as the study is quick to point out, you have to make it worth your followers' while, because "simply inserting garbage isn't going to lead to more retweets." Interestingly enough, hashtags weren't that effective in getting retweets, and @-mentions to other users had a marked negative effect.
• They followed their Twitter tribe. You don't want to stand too far out from your Twitter flock, according to the study. Users whose tweets conformed to what they typically tweet and what their community expects from them were "more easily accepted and therefore shared."

Other findings included imitating headline structures and using proper nouns. Now you can bring Cornell and Google's findings to your own social media posts. Along with the aforementioned quiz, the researchers also created a tool that lets you test which of your two similar tweets will be retweeted more.

Read more articles on social media.

Photo: Getty Images