Want to damage your brand? Then act on Social Media Analytics without understanding your data!

Social media is like that extra pair of remote ears you always wanted, enabling you to listen to and engage with the “word on the street” from your customers, potential customers and heaven forbid, your detractors alike. Properly analysed, the insights based on these interactions present you with opportunities not only to respond with the purpose of structured brand building and reputation management, but they also enable you to be proactive about designing products and services in response to sometimes quite clearly emerging trends before your competitors can. In other words, the opportunities for success are huge, but quite frankly, so are the opportunities for disaster.

There are probably just as many fantastic free analytical tools out there as there are paid-for tools, all depending on the same thing for their operation: data! This data is collected from as few as one to many different social media platforms, from blogs to Twitter, and are analysed in various ways from simple word or phrase counts, to sentiment analysis, either in relation only to the brand being considered, or more frequently in comparison to a defined competitor set.

Danger derives from the fact that few marketers know that data is where the problem of interpretation starts. It does NOT start with the output of analysis! That is because if your data is dodgy, then your analysis is just as dodgy - Garbage In, Garbage Out. The same holds for social media analytics. If you don’t understand the make-up of the data, then your interpretation could be way off, and even worse, you could then end up making decisions based on incorrect data that could be disastrous not only for your business, but for your entire brand!

Just one of many considerations required to ensure that your data is appropriate to make decisions from, is lack of bias. Your data should be spread reasonably equally across all the social media sources, being those individuals engaging with you represented by the lowercase letters in the examples below. In other words, no source should have volumes of comments significantly different in size compared to any of the other sources.

A borderline sample would for example appear as per Figure 1 below. All good analytics software, free or otherwise, provides either this information or the raw data required to be able to perform your own analysis.

Figure 1: Sources e and n are quite heavy over the selected timeframe given their weighting of 28%  between just two sources, and would probably provide a worst case example of acceptable data for analysis. Own represents the volume of your own interactions in the selected social media space

On the other hand, a distinctly bad sample would be per Figure 2 below, where your own social media interactions crowd out those of your own followers. This is not great, because the analysis would then be biased towards your own interactions, WHICH IS THE LAST THING YOU WOULD MAKE A DECISION ON! This could skew everything from keywords to sentiment. Remember, the critical thing about social media is the ears, not the mouth!

Figure 2: While all your data sources hover nicely between 1% and 5% of volume, your own social media interactions - 67% of the 5687 measured over the selected timeframe - totally crowd out those of your followers. Any analysis would be biased towards your own interactions, which are clearly undesirable in an analysis context! 

It seems obvious, but marketers ignore the basic rules of data, making me wonder whether marketers shouldn’t stick to marketing, leaving analytics to analytics types.