Twitter, Google and a social influence metric

[BenFox]

So we know that Google uses social signals from Twitter as a ranking factor but I was wondering if anyone had seen any research into the complexity of the social influence metric - the metric which differentiates between an authorative tweeter and a non-authorative tweeter?

Does anyone know if social influence is an iterative metric smiilar to PR or if it's calculated using the raw number of followers?

I'm a latecomer to this so if anyone can point me in the direction of information about what kind of data is contained within thr "Twitter firehose" it would also be much appreciated.

[tedster]

I haven't heard of any research - and if there is any it's likely to be quite premature, since this area is just emerging and I expect its evolution will be ongoing and the changes frequent.

One thing we can be sure of - Google will work hard to make social influence difficult to spam. With the prevalence of bots already on Tweeter, that's imperative. We can get some idea of what the complexity will be like by looking at other influence measurements such as Klout. It will be qualitative as well as quantitative - in other words WHO follows you on Twitter will mean a lot. WHO re-tweets a link means more than the mere fact that it is re-tweeted. And any influence metric is likely to be iterative, something like PageRank.

The Twitter Firehose data is a raw record of ALL Tweets and the user who tweeted, whether it was a Reply and so on. That means a lot because the more common Twitter "streaming products" only cover something like 30% of all tweets. I also understand that the Firehose feed does not indicate links as nofollow.

[tedster]

Along these lines, I have been noticing that even relatively low levels of mention activity on Twitter can trigger the live "scrolling" real-time box for an organic search on some user names. I first noticed this effect during the November Pubcon in Vegas. The scrolling box can disappear as fast as it appeared as mentions die off, and it doesn't necessarily appear for every user.

[BenFox]

Thanks for that - I figured it would still be early days.

Knowing there's that much data in the firehose is reassuring - less chance of one of my competitors gaming the system if Google's got a bird's eye view on it.

And regarding the scrolling box - we managed to get it appearing on a relatively competitive UK search term with <20 retweets. It does vanish very quickly though.

[FranticFish]

As well as Klout you might also want to check out the 'Monitor' Wildfire App, Trackur, and SocialMention.

Twitter have confirmed that they calculate influence for each user and use this to power the 'Who to Follow' feature and that they may even make this score public at some point in the future.

[Alex_TJ]

I think focussing on the number of followers is a bit too simplistic - it's more of a 'symptom' than a cause of social influence.
If I were building a 'twitter pagerank' I would focus on RTs, and maybe even incorporate replies if statistically significant.
Drawing parallels from websites, the username is the domain name, each tweet is a page, followers are just the number of visitors, and RTs are the links to that website.
The more RTs that user gets, the more authorative they become, and so their RTs of other users are also more authoritive, and so on in iterations.
Unfortunately I don't have the chance and resources to try it out though :(
Regarding the scrolling box, I'd assume it relies less on the absolute number of mentions, but more on the first derivative of mentions for that keyword.
Alex

[AlyssaS]

One interesting thing I discovered was that if you copy a bit.ly link into your browser and add a + at the end, it goes straight to the bit.ly stats page for that link - tells you how many people have clicked on it, who the referrers are - whether they came from twitter, email or facebook, how many facebook likes it got and so on.

If this stuff is visible to all of us, then it is definitely visible to Google, which means they can use the data and evidence of genuine sharing to influence which links posted on twitter they should pay attention to and which to ignore. I would imagine that the referral data on links shared through email which are actually clicked upon, are especially interesting to them, as those tend to be genuinely social.