How To Influence On Twitter: Research Results from New Algorithm Give Guidance
Recent work done at HP Labs, the exploratory and research group for Hewlett Packard, shows what most of us suspected as being true all all along; that just because a person has a lot of followers, it doesn’t mean they have a lot of influence.
In September 2009, using an algorithm they devised called the IP (Influence/Passivity) algorithm, a team of researchers from HP Labs continuously queried the Twitter Search API for 300 straight hours for all tweets containing the string of letters 'http'. Finding this string in a tweet would indicate the presence of a URL, and demonstrate that a web page was being shared or retweeted by means of a link.
In that time period, they acquired 22 million tweets with URLs present. This accounted for 1/15th of the entire activity of Twitter at the time. The URLs were checked for validity, and by revisiting the Twitter API they could determine who the user for each URL was, and in particular who their followers and followees were as well. From that information, a complete social graph was constructed from the dataset generated by the users sampled.
The research team worked on the following assumptions which are taken from their report "Inﬂuence and Passivity in Social Media":
- A user’s influence score depends on the number of people she influences as well as their passivity.
- A user’s influence score depends on how dedicated the people she influences are. Dedication is measured by the amount of attention a user pays to a given user as compared to everyone else.
- A user’s passivity score depends on the influence of those who she’s exposed to but not influenced by.
- A user’s passivity score depends on how much she rejects other user’s influence compared to everyone else.
A whole industry has grown up around Twitter with the aim of developing various tools that enable Twitter users to increase their number of followers. But now all these efforts seem to have been in vain. An average Twitter user retweets only one in 318 URLs. It seems most users are passive information consumers, and do not forward the content to the network at any kind of rate that could be described as 'little more than partially engaged'. Consequently, having a large follower count is not a lot of use from a message propagation perspective if most of the followers are made up of these passive users.
If you want to be a person of influence on Twitter, then the way to do it is to acquire engaged followers who are themselves active on Twitter. That would at the very least mean being active and engaged yourself.
Of course, this makes life difficult for marketers and others engaged in viral activity who want to take advantage of the enormous reach that Twitter has. They can no longer rely on a single dubious metric, follower count, as a guide to how far their message gets out.
It also means that to find active and engaged people, they will have to become active and engaged themselves. Fun, maybe. Time-consuming, certainly. But it is only through interacting with highly connected people that they will be able to propagate themselves and their message through the social network.
Through the process of finding the most influential people on Twitter, the team also managed to turn up the most passive users on the service as well. The majority of these users tended to be spammers and robot users.
It is as important to identify the highly-passive Twitter user because “they provide a barrier to propagation that is often hard to overcome.” It’s good to know where the Twitter dead ends are as it gives us a useful benchmark to contrast with someone who is influential. This information aids navigation through the vastness of the Twitter network, and knowing where not to go can be every bit as useful as knowing where to go.
The HP Lab report finishes with the following conclusion:
“This study shows that the correlation between popularity and influence is weaker than it might be expected. This is a reflection of the fact that for information to propagate in a network, individuals need to forward it to the other members, thus having to actively engage rather than passively read it and cease to act on it. Moreover, since our measure of influence is not specific to Twitter it is applicable to many other social networks. This opens the possibility of discovering influential individuals within a network which can on average have a further reach than others in the same medium regardless of their popularity.”
In a way, it is understandable that the findings from this research on Twitter would be applicable to other social networks as well. Influential people tend to be influential wherever they are. The IP algorithm that has been developed by the HP Labs team is going to be very useful across all of the social media domains, wherever people gather and exchange ideas and news via links.
The key metric to determining how effective any given person is in propagating information is to measure how often a URL that they tweet or retweet is clicked. And it is important to allow for the fact that not everyone is very adept at giving credit for their shared links, and some shared links go beyond the Twitterverse on to other services.
But unless you have access to a given individual’s bit.ly account or other some search service which keeps tabs on retweets, the only sure way to know if they are a person of influence is to get to know them.
BlogTalk is returning to Ireland on 26-27 August; check out the speaker list.