The proliferation of Twitter analytics tools continues. One of the most popular categories are tools that purport to calculate Twitter influence, that is, identify the individuals who influence others. They go by clever names – Tweetlevel, Twinfluence and Twittergrader to mention a few. (Here is a good list of Twitter tools from Brian Solis). They are fun – who doesn’t want to punch in their Twitter user name and see how “influential” you are? But are they truly useful? Should anyone base serious marketing or business decisions on these tools?
All these tools take a similar conceptual approach:
- Define the terms that are believed to be related to Influence.
- Arrange the terms in a sophisticated looking formula. Better yet – call it an algorithm.
- Factor the terms of the formula according to some proprietary coefficient weighting. (Tip: Labeling them ‘proprietary’ heightens the perception of mathematical precision.)
- Add a lot of terms to the formula so it seems complicated to the casual observer. (Tip: Complexity will heighten believability. Make it too understandable and believability may suffer.)
The game here is to create an aura of rigor when one in fact does not exist.
Equations are easy, coefficients are hard
I suspect in all the models discussed or available, the critical weighting of variables – assigning beta coefficients – was done by judgment, not math. To correctly assign coefficients, one would use statistical techniques involving means and standard deviations to determine the coefficient of each independent variable (number of followers, how often content is re-tweeted, etc.) and determine the relationships (correlation) to our dependant variable – Influence. The dependant variable should be observable and measurable. Here’s where it further breaks down. The problem here is no one is actually measuring true Influence – the ability of one individual to change another’s opinions, attitudes or behavior. You can’t surmise whether or not an opinion or attitude has been impacted, you have to conduct research. Opinions and attitudes exist within individuals. You cannot assess this by proxy, looking strictly at online metrics. Online behavior can be measured without primary research, but offline behaviors have to be observed or reported.
Influence is contextual
Influence is contextual not absolute. An individual may have the ability to influence certain people in specific subject areas. Authority and trust are important constituent elements of influence. Do they have the authority to speak within a particular area and are their words and deeds trusted? The notion of coming up with an influence score without context is inherently flawed. It might be interesting, but it is not actionable.
According to the results generated by this class of tools, I believe they are probably assessing popularity much more than Influence in a meaningful way. According to Tweetlevel this morning, Ashton Kutcher is the second most influential person on Twitter. Who exactly does he influence and in which areas? Mr. Kutcher is popular (number one in Popularity) but I’m skeptical of his true influence. He is also the most trusted person on Twitter according to Tweetlevel. The second most trusted person – Perez Hilton. Enough said.
Marketing not math
While I have been critical of these tools, I am not naïve enough to believe the intent was to create a rigorous analytic tool that could be used to target individuals that might have the most influence over your target customers. These tools are most likely designed to put a hand on your wallet, not insights in your marketing effort. Do they work for marketing purposes? Hard to say, but I’m sure they do in some cases. But, proceed with caution. You are walking a slippery slope in my view if you believe that developing a pseudo algorithm and slick website is the best indication that a company has the digital chops and experience to help drive your social business efforts. Getting to know the individuals involved and the work they have performed for companies like yours is preferable. Beware of shiny objects, they are not always as they seem.
metricsman.wordpress.com 
Recent Comments