You don’t have to move in specialist digital marketing circles to have come across the notion of ‘social media analytics’. Anyone on the receiving end of a ‘big data’ presentation from mainstream IT vendors such as IBM, EMC, HP, Microsoft, SAP and others will have had analysis of social media feeds put to them as a use-case for high volume number crunching technologies. Vendors that are particularly into digital marketing solutions such as IBM and Adobe also put a lot of emphasis on social media analytics in that context.
As I have sat and listened to more pitches, however, it has begun to concern me that presenters talk as if harvesting and analysing social media data is synonymous with tapping into popular opinion. The notion of using Twitter feeds to figure out what your customers and prospects are thinking and how they are acting is one I have heard expressed on quite a few occasions.
The problem is that no one ever seems to address the question of sample bias, and how representative any of this data is of the general population or group you are trying to understand.
But how can social media data be biased when things like Twitter are so open and broadly accessible? Surely the whole point is to allow everyone to express themselves freely.
What’s easy to overlook, however, is human nature. People’s inherent behaviour doesn’t change that easily or quickly just because you give them a new way to communicate.
Most of us will already be familiar with the concept of the ‘vocal minority’ versus the ‘silent majority’. We see this in action in various contexts. Whether it’s people sitting around a table in a business meeting, gathered together in a workshop or focus group, or making up the audience at a presentation or event, interaction often ends up being dominated by a small number of individuals unless other participants are actively prompted by the chairman, moderator or presenter. And when the naturally silent majority is prompted, it so often becomes obvious that the views of the vocal minority are not representative of the broader group.
In social media networks, especially those that have more of a broadcast flavour to them such as Twitter, the dominance of the vocal minority goes largely unchecked because there’s no equivalent of a chairman or moderator to ‘pull out’ alternative input. The views and sentiments you harvest are therefore highly unlikely to be representative.
Beyond this, the composition of the vocal minority is also interesting to consider. One group that’s always likely to be over-represented is the ‘naturally noisy’. These are the guys who will have an opinion on anything and everything, regardless of whether they have any direct experience or informed views on the topic being discussed. There are then the enthusiasts, evangelists and individuals with a vested interest who clearly use social media to promote their cause. At the other extreme, the ‘wronged’, ‘disappointed’, religious or philosophical objectors, and others with strong negative feelings are also going to be part of the vocal minority group.
The other factor to consider is that what people say (or don’t say) in public doesn’t always correlate with what they think and do privately. This can be for all kinds of reasons – desire to look different, conform, please, impress, etc, or fear of criticism, discrimination, persecution, and so on. This is why anonymity is so important to fair and effective elections in a democracy if you want the end to result to reflect the real wishes of ‘the people’.
The lesson from all of this is not to fall for some of the rhetoric we hear from IT vendors and digital marketing consultants about social media analytics being the be all and end all of getting to know your customer and prospect base. I would not for a minute suggest that anyone ignores social media as an important phenomenon, particularly with regard to customer influence (which is a separate conversation), but as a source of intelligence, it can be dangerously misleading if used in isolation without appropriate corroboration and/or correction for bias.
From a systems perspective, this brings us back to the importance of social analytics systems, and indeed big data solutions in general, being properly integrated into your broader landscape of analysis, planning and decision-making capability so you can work on the basis of a more holistic view of the customer.