The notion of having your employees tweeting on your behalf is indeed a powerful one, and a recent MIT article looks at the potential power of having an army of brand advocates set loose on social media.
Of course, such an endeavor is not without risks, as I’m sure anyone connected with UKIP could tell you. You do open yourself up to people saying stupid things in your name.
A new Cornell study might help companies to inform their staff on the right kind of things to be tweeting. The research team have developed an algorithm that they believe can predict which version of a tweet will be shared more often.
A website has been created to allow people to conduct pre-emptive A/B testing on their tweets to run them off against one another.The study used real Twitter data, with specific steps taken to remove any possible impact of the users popularity on affairs, or indeed the popularity of the topic itself.
The algorithm was trained by comparing thousands of pairs to try and understand what it was that made a tweet sharable. It’s especially important when the tweet contains a link that the user hopes people will click on.
“You want to say something about it to make people look at it,” the authors say.
Of course, there have been several other studies into what encourages us to share a tweet (check them out here and here). The researchers used these findings to set them off against each other before assimilating the findings into the algorithm.
It looks for particular keywords and ‘bigrams’ or combinations of words. These combinations may be a reflection of the linguistic style of the user.
They ended up honing in on nine things that often resulted in more retweets.
- Asking people to share the content
- Being informative
- Using the language used in your community
- Mimic headlines used by newspapers
- Copy words that are used in other oft. retweeted messages
- Take advantage of emotive words, whether positive or negative
- Reference other people
- Use statements that can be applied to many situations
- Use relatively straightforward language
It’s noticeable that there isn’t really any account taken of humor in the results, which the authors admit was something they struggled to capture.
“We would love to capture amusingness or cleverness, but we haven’t found a way to do that yet,” they say.
The results aren’t massively different to the findings produced by previous studies, but it is quite nice to have a tool whereby you can compare the share-ability of your tweets.
Check it out and see what you think.