Stephen Waddington on “Tackling Twitter tarts”

Posted by Stephen Waddington
on 27th April 2011
Bookmark and Share

Twitter tartMonitoring and measuring influence in a social network is a tricky business. It’s little wonder that it has spawned an army of analytics and measurement firms.

Measuring Twitter influence

When it comes to Twitter there are two major players in town, namely Klout and PeerIndex, each using its own bespoke algorithm to plot the authority or influence of a Twitter user.

Marketing and PR professionals seek out the most influential people in a market segment as the target of their campaigns.

Meanwhile consumer organisations have yet to add social media profiles to their CRM systems and so being able to organise customer complaints by relative influence is useful.

Networking is easy

But what is the value of these tools when Twitter can be so easily manipulated?

A recent audit of a series of business-to-business Twitter accounts by my own firm Speed found that more than 30 per cent of followers were spammers or unused accounts.

We’ve already seen the emergence of offshore businesses that will build networks for you promising 2,000, 5,000 or 10,000 followers within 30 to 60 days for a couple of hundred dollars.

If you want to play the numbers game grab one of the Twitter tools that allow you to create a pipeline of target followers based on a keyword, piggyback similar accounts, and automatically unfollow people that don’t follow you back.

Accounts that have been built in this way are easy to spot. New followers will have been added uniformly over time and followers and following numbers are likely to be similar.

There is an even simpler way to add followers quickly. Tweet more and you’ll get more followers. The relationship between the frequency of tweets and the number of Twitter followers is well proven. But on Twitter, as in life, the noisiest people are unlikely to be the most influential.

Abuse goes untracked. To date Twitter’s action against anyone that abuses the network has been limited.

PeerIndex tackles Twitter gaming

I caught up with PeerIndex’s Head of Products Simon Cast to discuss this issue. He is responsible for developing the PeerIndex algorithm and recognises the problem.

“PeerIndex spots outright bots and while the number of followers plays a role in the calculation of a PeerIndex score it is not as big as you might expect,” said Cast.

PeerIndex addresses the fact that having lots of followers isn’t necessarily an indication of authority. The audience must listen and be receptive.

“Tweeting content that resonates with your network is one of the most important metrics of influence, and retweets and replies are the best measure of this,” said Cast.

Accounts with lots of followers that aren’t engaging with their network are already weighted appropriately according to Cast and changes to the PeerIndex algorithm in the coming few weeks will counter this issue.

“People will always game a system. As quickly as an algorithm is modified users will find ways to manipulate it for their own end,” said Cast.

This is the game that Google has been playing for the last 15 years constantly hardening its search algorithms to counter abuse. We shouldn’t be surprised that history is repeating itself on Twitter. But companies such as PeerIndex are making a bid to keep the network honest.

Stephen Waddington is md at Speed Communications. Image via SeattleClouds.com

Tags: , , , ,

Recent comments
  • trufflenet

    If influence can't be measured in the 'real' world then it can't be measured in the digital sphere and vice versa. Far better to find a more definitive word and measure for that than to aim for the impossible with an algorithm. And the answer surely lies in calculating lots of different things.

  • As Andrew says, "What is influence?" is a key question - after all, according to both Klout and PeerIndex I rank higher than the Deputy Prime Minister. Yet he has rather more power, influence, media power, staff, budget and so on than me - and even in the online world alone, what he does gets far more coverage and generates far more interest than my own activities.

    That's not to say they're without uses as measuring tools, but you have to be very careful what you use them for (more on this at http://www.mhpc.com/blog/nick-... )

  • I think there are a whole load of issues entwined here - the big questions are: what is influence? Can it be objectively identified and measured? And can certain people or organisations use this as the basis to claim that they can then measurably impact how influence is exerted?

    Surely there is a difference between conscious and and unconscious influence. And as I keep banging on about it, influence is always relative to something or a context. I have an influence on what technology we buy in our household. But I have no influence on what clothes my wife buys. The network topology of a group of people has a bearing on things - the same group of people may help spread a disease, but not be receptive a to PR message.

    On a pedantic point, does PI claim to have anything to do with influence? They talk about authority - and as I've said before, there is clearly something in their attempts to link the PI score relative to a particular topic rather than a generic score.

    Of course, you are absolutely right to say that this isn't simple. I realise PI is by no means perfect - but I do think there is something in the idea of taking a Google-style approach to the attempt to analyse authority or influence. However, it needs to be combined with a whole host of other things such as network topology analysis, etc - we have barely scratched the surface.

    Suffice to say this one will run and run. Not least because I need to finish reading your book - which I will probably require me to write a book in response ;-)

  • I think Peerindex have been very active and responsive in changing their algo. The PR Week Power Players of Social Media list that I curate is a case in point: http://bit.ly/giceIS

    I know some people questioned why certain people were ranking so highly in the list a few weeks ago ie was it simply due to the number of followers they had or volume of Tweets (ie automated). However, looking at the list today, you can see that PI have clearly made allowances for that.

    As I've always maintained, influence is always relative to a subject or topic. Looking forward to PI rolling out its topic related ranking system fully.

  • Stephen, in my opinion, Klout and PeerIndex have nothing to do with influence, and they should be ashamed to claim as much without empircal evidence. Their interesting network science aside, they confuse popularity for influence; they confuse expertise for influence.

    Rather than second-guessing what may or may not have influence, we should relegate such specious measures and work with facts. What has exerted influence?

    You have been influenced when you think something you wouldn't otherwise have thought, or do something you wouldn't otherwise have done. If neither of these are manifest, there has been zilch influence.

    Ultimately, if rather controversially, I think marketing and PR practitioners are groping for a simple world like the one they apparently left behind last century. They want it to be simple. They want a list of five / a dozen / two dozen "influentials" so they can tell the client / boss that they're on to it, and feel good as they tick each one off their list. Lunch with _, tick. Phone conversation with _, tick. etc.

    But influence has, is and always will be considerably more complex. Empirical studies show that most of us are more influenced more often by the 150 nearest and dearest than the other six billion people combined. And the sooner marketing and PR practitioners learn how to approach such complexity, the sooner they will be able to demonstrate tangible outcomes rather than specious outputs, and win the authority and accountability that comes with a place round the board table.

blog comments powered by Disqus