SIS INTERNATIONAL RESEARCH: NYC FOCUS GROUP FACILITY
Unearth the power of declared data
Insights That Work

Social Analytics Traction… At Last

Social media analytics appears to finally be on the path of exponential growth forecast for it. A number of reasons account for why growth has been slower than initially expected, and why growth is picking up now.

Editor’s Note: The latest GRIT findings, previewed in these pages earlier this month, show that use of social media analytics continues to grow, especially among client-side practitioners.  Despite this continuing growth, optimistic forecasts from earlier this decade suggested that its use would be even more widespread by this point.  Michalis Michael discusses some of the reasons for this state, and argues that growth has turned a corner.  The advancements in the science underpinning social analytics have been key.  Given these, can we finally consider social media analytics mainstream?


I think every market research professional will agree, that it is unthinkable for a company to be selling products in major retailer chains and not be tracking sales, volume and value shares, pricing and distribution, by subscribing to a retail measurement report on the relevant product category. Isn’t it therefore only natural that social media listening reports ‒ for brand share of voice ‒ will one day be as important as retail measurement reports ‒ for brand market share ‒ to the Marketing Director of a CPG company?

It took a bit longer than expected but 2018 was a good year in terms of progress. Fellow agencies and end clients alike started to dabble in online data sources and high precision analytics using machine learning, on the path to insights discovery, campaign evaluation, brand reputation management, even micro-influencer identification or lead generation. Social media listening & analytics traction… at last!

Humans tend to think in a linear way. We also tend to overestimate the short term and as a result we underestimate the long term. Case in point, I “bought” Joan Lewis’ (then SVP of Global Consumer and Market Knowledge at P&G) prediction when at a 2011 ARF conference she said:

“Survey research will decline dramatically in importance by 2020, with social media listening replacing much of it and adding new dimensions.”

What I extrapolated from this statement back then, was that by 2020 at least 30%-40% of the market research budgets would be spent on social listening & analytics, moving away from surveys and other traditional MR methods. So here we are today, just under two years to go until the end of 2020, and with the social analytics market being – according to Reuters – only US$3.4 Billion in 2017. The total market research market was US$76 Billion in the same year; social listening accounting for a mere 4.5%.

Maybe you have heard of the 10-years-to-overnight-success pattern; it became a thing after Jeff Bezos said it for the first time; it looks like social analytics may be another case to prove the pattern. In the graph below you can see the flat linear growth between 2010 and 2015 and after that, the 2020 forecast for US$9 Billion (by DigitalMR in January 2017) and the 2023 forecast for US$16 Billion (by Reuters in February 2018). It looks like we are finally on a path to the exponential growth we’ve been expecting year after year.

 

 

There are many reasons the market research industry has been so slow to embrace social listening & analytics, and fear of the unknown may actually be a part of it, but I think it was mainly due to:

  1. Loss of trust ‒ buyers of market research tried social media monitoring tools developed by pure technology companies a few years back, they found the accuracy of the analysis to be low and decided this approach is simply inappropriate for the purpose of market research.
  2. Existing subscriptions ‒ client-side market research departments are sometimes using the social media monitoring tool their media agency or digital marketing department subscribes to, in a very superficial way as they believe that this discipline sits somewhere else – in another silo. As in point 1 they believe that it has limited applicability to them as insights professionals.
  3. Where are the insights? ‒ whether ad-hoc reports from their media agency or analysis received directly from the aforementioned DIY tools, MR buyers have been left underwhelmed in the past with the reports from social – even if we ignore the accuracy problems. Technology can only go so far, you need someone with consumer insights expertise to make the most out of the analysed posts.
  4. Let’s stick to what we know ‒ potentially related to the individual’s long experience and years spent working with traditional MR methods, there are still colleagues out there who refuse to let go of notions such as “sampling” and “respondents”; they find it too risky to accept social intelligence as a new and valid market research method that can complement and sometimes completely replace existing methodologies.

Most social media monitoring tools can achieve 60% accuracy at best (many way lower than that) on any of the important metrics i.e. brand relevance, sentiment or topic. This is not due to inexperience or lack of knowledge or options, but purely a choice made by companies to maximise profits. It actually takes a lot of effort to create custom machine learning models ‒ which is still the only way to reach accuracies of over 80% if done right ‒ but the extra effort pays off in the end. Even machine learning can get it wrong if the model is trained on low quality (or the wrong) data! In a 2013 face-off of market research versus four traditional social media monitoring tools, with Carlsberg as the client, we found that over 90% of the online posts analysed by the four were not even about the brand and the digital campaign which was to be evaluated. In another face-off for SABMiller, 80% of posts classified by the social media monitoring tool as negative were found to be positive, and 56% of the posts classified as neutral were in fact negative. A third party appointed by the client found the overall sentiment accuracy of the market research approach to be 87% vs 44% for the social media monitoring tool approach.

Growth will only continue to happen if the social listening and analytics vendors out there can prove to non-users that there is real value to their business, by demonstrating that they can accurately interpret what customers are saying on social media and translate it into actionable insights. On the buyer side of things, there needs to be more open dialogue between the various departments of an organisation. There needs to be an alignment of interests across the business as a whole, and a realisation that everyone can benefit from such solutions more or less equally, just in a different way. Social listening & analytics is not just for the marketing department, it’s not just for PR crises, and not only about monitoring the success of corporate social media posts; social insights can serve many purposes, and a “one stop shop” that works for all departments is better for the company as a whole.

Beyond ‘social’ as a standalone source of data and insights, there is ample evidence that integrating social listening with surveys and real consumer behaviour such as purchases or website visits brings impressive value to the table, particularly when it comes to discovering correlations between social sentiment and sales. As Schweidel and Moe also suggest in their paper Listening in on Social Media: A Joint Model of Sentiment and Venue Format Choice, published in the American Marketing Association’s Journal of Marketing Research in 2014, “examination of this sentiment could provide guidance to brands on which domains are most critical to monitor and actively engage social media contributors”. In a R&D study conducted in 2017 and presented at the ESOMAR MENAP conference, it was impressive to discover that the beta coefficient for positive sentiment and sales was double than that of negative sentiment and sales! Of course, all the effort and resource you put in social analytics and data integration is a complete waste if the relevance of the data and the accuracy of annotating the posts with sentiment and topic of conversation is lower than it should be.

Progress has been painfully slow up to now, and according to the forecasts described above “listening” will only account for around 20% of the total market research market by 2023, but hey, much better than the 4.5% it was in 2017.

 

Please share...

One response to “Social Analytics Traction… At Last

  1. agree with the sentiment! Custom models are very important as language differs by industry. There are new solutions now that make these models easily available. And, yes, its important that brands know the actual performance of their models before deployment (f1 score a minimum) and avoid any inadvertent bias. Nice write. up.

Join the conversation