HP aims to measure Twitter influence
Researchers find the most influential Twitterers aren't the most prolific or most followed, but those whose tweets bounce across the Web like a rock skipping across a lake.
TRUCKEE, Calif.--So what makes someone on Twitter influential?
My two cents is that it starts with not posting your every Foursquare check-in, obliquely mentioning meetings you can't talk about, or sharing your latest bodily function.
But a team at Hewlett-Packard Labs tried to find a more scientific answer by analyzing 22 million tweets published in a short span. It found that it's not the visible metrics that truly define the influentials.
Rather, influence is better measured by those whose tweets spread far and wide--something that is not so correlated as one might think to the number of followers that a particular person has.
<="" p="">"Most content goes very few hops," said HP Labs social-computing director Bernardo Huberman, in a meeting over lunch at the Techonomy conference here. It's the latest report from Huberman and team, who have also studied the best time to post on Digg and demonstrated how Twitter can be used to predict a film's box office success.
Huberman also has bad news for folks who think posting a lot is boosting their influence.
"I wouldn't call you influential, I would call you energetic," he said.
So, it seems the key is not just having followers, but having active ones that like to share your thoughts as opposed to those who just read. Having something worth saying probably helps, too, but that was not the subject of HP's study.
Why it matters, beyond perhaps helping me in my vain quest to crack 10,000 Twitter followers, is that the deluge of information means that there is fierce competition for issues seeking attention.
"We only talk about things that bubble to the top," he said.
Of course, identifying influential people also has other uses, such as telling companies which bloggers and tweeters to target or governments and nonprofits where their key audiences are.
The full research, conducted by Huberman and colleagues Daniel Romero, Wojciech Galuba, and Sitaram Asur is published on Scribd. You can read the whole thing after the break.