Monday, March 26, 2012

Is online influence measurable or meaningful?

Online influence is a hot topic right now, with companies such as Klout, PeerIndex, Empire Avenue and Kred all building online services that aim to measure the influence of internet users, in order to better target advertising dollars.

But how effective are these services really?

Does the number of followers, retweets or likes or some form of combination really identify those most likely to influence decisions and behaviours on a large scale?

Would any of these services have identified Janis Krums as an influencer of millions, before he tweeted a photo and message to his 170 Twitter followers about the plane that had landed on the Hudson River?

Would they have identified QLD Police Media as an important and influential account a few weeks before the Brisbane floods?

Would any of them have identified Rebecca Black, singer of 'Friday', as influencing an entire generation?

Influence online can ebb and flow rapidly. People go from virtually unknown to globally famous to unknown in a matter of weeks, days - even hours.

Therefore I was interested, but perplexed when I received the following email from PeerIndex a few days ago.

PeerIndex email:
I work at PeerIndex and we have a group on Australia top Twitter influencers and was wondering if I could get your feedback because you are on the list. PeerIndex measures interactions across the web to help people understand their impact in social media.

I was wondering if you could look over the list and let me know if you felt it was accurate? Do you recognise the other people on this list?  Is it missing people that you think are important?  

We would like to open up a dialogue with people in your field and think this would be useful to them (or at least start a conversation) it was accurate and interesting.

Thanks very much for your time,

I had a bit of a think about this and realised that I am an influence sceptic.

I am interested in sentiment online - whether people believe/perceive and say good or bad things about a topic. I think there's a strong future in this as a way to judge a general mood, supported by other more refined techniques.

However influence is just too hard to measure if only one dimension - online is taken into account.

Hence my reply, below:
Hi ,
I would love to help, however I don't think I honestly can.

I just do not understand how influence on Twitter, or on other online or offline social networks or situations, can be calculated in any effective manner.

Interactions online don't necessarily translate into actions offline and influence is generally a subtle and cumulative process - which requires multiple sources over a period of time.

For example, you tell me something on Twitter, I see something related from someone else in a forum, it gets discussed at work, I do some research as my interest is raised, then it appears in the traditional media and then I see others I trust taking a position and then I do.

The interlockings between topics and influence are incredibly complex and related to individual mental models and worldviews. Something that would influence one person will have no impact on another, people weight influence based on source, channel, frequency and relationship - and every individual has their own influence model - what will or will not change their view.

For an example (or study) of this, just watch the classic movie '12 angry men'. It is a brilliant look at how varied the influencers for different people may be.

I don't think there is a reliable way to identify influencers or put people in boxes for influence.

I find your, and other similar services, amusing, but do not see how your algorithms have accurately modeled my, or anyone else's levels of influence on the micro or metro topical level. 

Your models are simply far too simple and work on a subset of observable influences with no characterization of the individual influentiability of different people in different environments at different times - nor how long-term that influence will be.

Behavioural psychology is an extremely complex and poorly understood science. About the only way we can reliable detect influencers at any specific time or micro topic is in hindsight.

Humans are lousy at determining what is likely to be influential, other than by 'gut instinct', or through sledgehammer techniques, such as mass repetition (show the same message enough times to a broad enough group of people and some will be influenced).

So sorry, I don't know what makes people influential - chance, chemistry, repetition, a match with a particular mental model, a combination of influencers all working in alignment, or a reaction against a 'negative influencer' (a de-influencer? Someone we love to disagree with).

I certainly don't see how dividing people into boxes by arbitrary topic helps define their broader influence, or specific influence across other topics. The amount they talk about a topic isn't a good judge either, and it is always unclear whether someone 'heard' the message on a service such as Twitter.

So I don't think I can help you. Nor am I sure if your service, or Klout or the others in the space has a real business model. Though I do hope that your collective efforts expand our understanding of how connections between people can sometimes influence them.



What do you think?.

1 comment:

  1. That was a polite response, and I'll be interested to see their answer.

    I've looked at services like Klout and while I can see the point, I'm not interested in their attitude to "influencing" and of course, how they measure it.

    The most significant thing about their mission (from their About Us page) is "We believe that marketers want to engage with exceptional Influencers". Be warned.