Dollywagon White Paper defining Social Web Influence

Analyse your online ecology

Cut through the tsunami of content delivered by your web monitoring system to find the most influential people who make the weather around your brand.

Take a tour

Manage your online ecology

If you want to inhibit the spread of bad brand perceptions whilst promoting positive ones, the Influence Engine is the most rigorous system available.

Take a tour

As the importance of the Social Web grows, the need to define and measure influence robustly becomes more urgent and widespread. But defining and measuring Social Web “influence” often seems like a black art.

A quick Google search will turf up many systems and ‘black box’ methodologies that claim to do the job. There’s a lot of innovation in this field and “Influence Ranking” is a hot topic.

Influence metrics are increasingly being used to inform everything from product development, customer support, public outreach, lead generation, market research and campaign measurement.

This underlines the growing importance of ‘Influence planning’ as a supplement to more traditional methods of consumer research and media targeting.
However, in our view the general trend of market development is failing to provide a robust and defensible consensus on the best way to approach the measurement of Social Web influence.

The key flaw with leading Social Influencer Rankings is that they do not derive true networks from the Social Web data they collect. This is crucial because a formidable body of scientific evidence (see Barabasi et al) demonstrates it is impossible to grasp the true influence of any ‘player’ within in a community if one does not take account of the cat’s-cradle of interpersonal relationships within which each person is embedded. This is known as the ‘social graph’.

Most ‘Buzz Metric’ or ‘Influence Score’ companies do not address this issue adequately. In general, they tend to measure influence by aggregating, in various proprietary ‘black box’ ways, various Social Web data.

These systems generally ‘count’ things. For instance, blog posts with lots of links, or people with lots of ‘friends’ etc. are assumed to be more popular. Blogs with lots of ‘inbound’ links, or Twitter users with lots of ‘followers’ and RTs (re-tweets) etc, are assumed to be more influential.

We call this ‘node-centric drop-down data’. In a very reductive sense it tells us a lot about each individual within a system. But it’s not possible to understand the dynamics of a community by looking at statistical data on how many links, positive mentions, RTs or followers a social web user has generated. This information is useful, but just ‘counting’ it is not enough.

Herein lies the main ‘philosophical’ concern about current trends in defining and measuring influence. Reductive methodologies tell us nothing about how neighbours within a community actually rub along together. It’s a bit like trying to understand the sound of an orchestra (i.e. the network) by listening to each musical instrument on its own. You can’t really ‘understand’ an orchestra until you’ve heard all of the instruments play together.

‘Influence Scores’ based on a reductive approach are nothing more than arbitrary statistical analyses of aggregated Social Web data. What’s really needed to measure influence is a more complete and holistic picture of relationships and interactions between members of a community. Which is where a second general approach to defining and measuring influence comes into the picture.

Social Network Analysis (SNA) is a fast growing scientific discipline that doesn’t just study the drop-down data associated with individuals social web users (although it does make good use of this data).

It also analyses the pattern of relationships that exist between people throughout a social network. To compute ‘influence’ robustly it’s necessary to collect more complex information about the ‘life’ of a community, such as:

• who-links-with-whom
• who-follows-whom
• who-responds-to-whom
• who-is-interested-in-what etc.

We call this ‘proximity data’, which provides the basis of a true social graph. The pattern of relationships or connections that surround each one of us effectively betrays our personal significance, credibility or influence within our own social networks.

Our connections rarely lie – they are often a sincere reflection of the significance or respect that individuals have for each other. We know this is true, even at an intuitive level, because people still get tetchy about things like peer group status, social class distinctions, ‘old boy networks’ and Masonic institutions.

Proximity data enables us to construct ‘topological’ network structures from the complex soup of relationships that exist in any Social Web community. When we analyse these relationships ‘Influence’ becomes a mathematically defined property that emerges naturally from the pattern of connections we establish with our online friends and acquaintances.

This is a difficult task that cannot be avoided if one is serious about measuring influence. In our view the field of Network Science offers a robust mathematical foundation and an open body of scientific literature that provides a basis for industry consensus on the most effective way to define and measure ‘influence’.

Experience in the media industry demonstrates that the development of any useful audience measurement methodology generally trends towards higher quality and more open standards. Such trends are naturally driven by the need for advertisers to make better informed media investment decisions.

Most conventional ‘Buzz Metric’ and ‘Influence Score’ approaches seem arbitrary and incapable of achieving a defensible measure of network influence that clients can embrace. If this exciting new industry is to thrive, it’s time to start talking seriously about Network Science and the benefits of its holistic approach to measuring Social Web influence.