Corporates tend to approach social media with a ‘social listening approach’ – looking for mentions, basic sentiment analysis and not much more. This is a low budget option that helps you stay abreast of your brand on social media.
Vibrand Opinion, though, takes a different approach. To us, social media is an ocean of valuable data that we can mine and analyse. Vibrand Opinion works in partnership with Crimson Hexagon, arguably the world’s leading social media analytics platform. We are able to access a database of 900 billion (yes – 900 000 000 000!) and counting pieces of social media data, dating back to 2010. Information is captured from Facebook, Twitter, Instagram, Tumblr, blog posts and forums. Every single public social media conversation globally has been captured and is available to us, and is reliable for analysis as far back as 2010.
Also, basic sentiment analysis relies on untrained algorithms, or a modicum of human intervention, and are often wildly inaccurate. When we train ‘monitors’ to look for posts and assign them to various categories there are analysts with years of experience trawling through thousands of pieces of data to input them into our ‘monitors’. They then use the most advanced algorithm imaginable (thanks to Crimson Hexagon) to analyse the rest of conversation around the topic. This to a 95% accuracy rate.
This presents an opportunity for almost every category and brand to mine social media opinion in any number of ways.
We’re not big fans of ‘products’ at Vibrand, as we prefer to approach each project on a case by case basis, but we’ve cobbled some ideas together – with a tool that can do anything though, we would prefer a brief that we can construct a detailed approach towards…
You may be thinking of having a celebrity endorse your brand, but you are unable to decide which to choose. Find out which schleb is mentioned the most; use our monitors, trained by experts to examine volume, sentiment and any nasty skeletons hiding in the social media closet. All the way back to 2010.
Affinities (interests) analysis is a very powerful tool, and can easily be added onto category and detailed sentiment analysis. Make no mistake, as we imbue brands with personality, certain personality types respond to your efforts. So personalities choose brand personalities. And as a result, those that use well-defined brands have different personalities and interests to those that use other well defined brands. Sounds fuzzy, but with affinities, we are able to see what the interests of various groups are. So let’s imagine you run a bank… We can isolate all consumers that speak favourably about your bank and all competitor banks. We can then look at each of their sets of interests against all of twitter, and against each other. In this way, we can construct a detailed ‘interest map’ – bank 1 might have consumers that like sport, white wine and swimming; bank 2 might have consumers that like finance, the stock market and David Cameron. Different banks, different personalities… This allows you to either retain your core market through your sponsorship and marketing strategies, or attempt to acquire consumers in your competitors market through the same.
All our products can be run retrospectively – theoretically to 2010, but more useably to 2013. So if we stick with affinities – we can run affinities on any set of consumers and track their behaviour over time. How have their affinities (interests) changed over the years – what is growing and what is declining? We can also do this with any category, brand and competitor data.
We can, for example, run affinities on age groups – under 17 vs 18-24 vs 25-35 in any country. We can see how their interests differ, and we can see how they have changed over the years.
Using affinities carefully (and laboriously) and co-analysing them with affinities run on related hashtags (long story) – we are able to conduct segmentation analysis. Here, we are able to break your market down, with a high degree of accuracy, into interest segments. Did you know, for example, that people that speak about germs on social media are 6 times more likely than the global average to be interested in parenting? They are also 3 times more interested in ‘being a mom’, 18 times more likely to be into coupons (value based) and a staggering 48 times more likely than the global average to be into talk radio… So, for my money, if I was making a product that claimed to kill germs, I would aim it at moms with young kids, I’d use a discounting strategy and I’d advertise on talk radio!
You want to know about your brand, but also your category – what jean cuts are the most popular; how do people feel about germs; how do people feel about cleaning their houses; what do people look for in a new car and so on and so on. All possible on a country by country, regional and even global basis.
You make a throat lozenge, and what to find out how people speak about sore throats; you make a chewing gum and you want to know how people speak about freshness. You make a lasagne and you want to know where and how people eat lasagne – alone or with others? Do they make their own or buy from a store? How do they describe a bad lasagne? And a good one?
You want to know more than just sentiment – you want to know how you and your competitors compare when discussed; what drives sentiment; how do people see you and your competitors
The apex of our work is correlation and regression analysis. So, this is complicated but here we go… Let’s be that guy that runs a bank again…
We construct monitors that very accurately reflect sentiment and volume – not just on a bank by bank basis, but on the interaction between the banks. We also include affinities, breakdown of volume for each bank (by province and by city), male/female breakdown, age breakdown, the ‘USPs’ of each bank and look at what drives sentiment for each bank. This can be taken back as far as 2010. It is then presented as a monthly report (on each bank). Then, once we have looked at what drives sentiment towards all the banks we construct a separate monitor. It evaluates, with no connection to banking at all, the main factors that influenced sentiment. In this case they may be technology, customer service, national pride etc. We construct this monitor to run over the same period as the various banks monitors ran.
Then we ‘grab’ the data on a day by day basis (yes, we can do that!). If we were to take data back three years, for example, had five banks we were investigating, had five ‘main factors that influence sentiment’ we would have 365 days x 5 banks x 5 sentiments (positive, advocate, negative, angry, disappointed), 5 factors x 3 years = 136 875 data points to look at. We then run a detailed correlation analysis – which looks for statistically significant correlations between all these data points. We look at those correlations and – because correlation does not imply causation - where they make sense we report on them. You start to build up a knowledge base of influences. So when technology is being spoken about, it has the following effect on all 5 banks; when customer service is being spoken about it has this effect and so on. Then, we work towards a regression hypothesis – look at what has happened before in order to predict what will happen in the future.
So our guy, who runs a bank, decides to sponsor a rugby team, or up weight his tech, or speak about his customer service - our regression analysis should be able to predict the response on social media to each of these efforts. This ultimately allows us to devise strategies, based on all we have found out about his bank and his competitors to retain his core market while acquiring competitor share.
We approach each project on a case-by-case basis and design an analytical strategy based on the objectives and the data we uncover. Research on Opinion is a fact finding mission – we can hold no pre-conceived notions about what is there, but what we can do is assure you that we have the best tools and the best possible analysts available to mine and make sense of relevant social media data.