It is truly a dangerous thought that the collectivity of our understanding of social media and the insights we draw from it are based on analytics platforms, data, and algorithms we know actually nothing about.
Social media platforms have created a realism distortion field that builds up them as the very explanation of “bid data ” while in existence the entire documents of every link shared on facebook and every tweet ever sent is extremely smaller than we might ever have fancied.
In fact, just a fraction of our daily journalistic output is almost as large as many of the datasets we work with.
In turn, the social media analytics platforms we swing to make an impression of social media are black boxes from which we madly report without ever asking whether any of the tendencies they give us are real.
Is it time to just give up on social media analytics?
One of the most amazing aspects of the social media analytics industry is just how little clarity there is into how any of these platforms work.
Chart sentiment, map user clusters, identify users and drill into demographics, customers plot volume timelines, compile author and link histograms, all without having the minor insight into whether any of those results are positive.
Over the past year I’ve had the affliction of having to use the results of a number of social media analytics platforms and my involvement have been exactly eye opening with reference to just how bad the social media analytics space has become.
Even some of the gigantic players make heavy use of sampling, use algorithms that have been generally established not to work exactly on tweets, apply sampling even in their documentation states especially are not sampled or make correct allegations regarding the accuracy of the data, methodologies and algorithms they use.
Most regarding, few platforms are up front as to the ramification of their multitude methodological and algorithmic decisions on the findings that their customers draw from their tools.
Their sleek web interfaces make no acknowledgement that results are sampled or that there was a detaching change in a key algorithm that will cause a colossal change in results.
In some cases, both their documentation and interfaces explicitly state that results are not samples, but after being encountered with authentic evidence, a platform will softly recognize that they do literally anticipate results and thus results may be automatically off or even absolutely wrong for certain displays.
If that is the case, what’s the point of practicing analytics platforms at all?
Putting this all together, for all their sleek interfaces and hyperbolic marketing materials touting precision analytics, the sad phenomenon is that many of social media analytics platforms out there today turnout results that are controversial at best and outright farcical at worst.
In the end, Instead of concentrating on packing every sensible features in their systems and believing efficiency comes secondary to speed, for social media analytics platform to sophisticate, they need to concentrate more on bringing accuracy first, even if that means spending a sample more on their computing infrastructures and compressing the number of features they offer.
After all, an elegant graph isn’t worth much if the story it tells is entirely wrong.