Brands have found the immediacy and immense reach of social media to be invaluable for a range of functions, in particular for listening to customers in real time or evaluating campaigns. But the vast volume of data that social media provides can do much more. Properly analysed and understood, it can play a major role in strategic issues that confront all brands, such as uncovering innovation opportunities, personalisation strategy and new avenues for monetisation.
However, certain characteristics of social media, such as skewed demographics and underrepresentation of internalised needs, can render social media, on their own, a rather unreliable foundation. On the other hand, brand strategy has typically been underpinned by foundational representative survey segmentation studies conducted every three to five years, a timescale dictated by the cost and time resource required. But in a world where the market, brand and consumer landscape are evolving at an everfaster pace, such a timescale for foundational study data is no longer fit for purpose.
This calls for integrating the scale, spontaneity and continuity of social media with the representativeness of survey insights to transform the way foundational insights are conducted today for strategic planning.
This article will illustrate, with the help of a study into quick service restaurants (QSRs) in the UK, how social media insights can be harnessed, together with survey insights, for developing strategic planning on a leaner, richer and continual scale. This will help marketers in developing personalised one to-one in-the-moment marketing strategies – much needed in today’s landscape of increased proliferation.
In April 2014, TNS conducted a selffunded survey into the QSR category in the UK, which yielded six need-state segments, based on a psychological segmentation framework. The following month, TNS mined Twitter data with regard to the QSR category in the UK. Of the 155,000+ conversations, only around 25% (37,559) conversations were related to meal experience and usable. This reflects on the absolute vs. relevant volume of social media data. However, 37,559 is a huge sample to analyse to provide a high level of granularity. So these 37,559 conversations were allocated to one of the survey segments based on an artificial intelligence (AI) algorithm.
Survey data provided an important benchmark, and analysis of the social media data indicated that some segments were less represented online. The most pronounced difference was noticed in the ‘Contentment’ segment (segment size of 2% from social data vs. 17% from representative survey data). This reinforces the role of survey data: you cannot establish foundational insights for strategic planning based on social media data alone.
The Contentment segment represented the basic first-order internal consumer need for trustworthy restaurants with a relaxed atmosphere to help feel at ease; it is not surprising that consumers seldom express such needs via social media. So if Contentment is your target segment, social media would not be your chosen marketing strategy, but you can still use social media data for generating insights for the segment, given the large volume of conversations. Despite the 2% size on social media, the Contentment segment had over 700 conversations in only one month’s data on Twitter alone.
The AI algorithm is based on knowledge of linguistics which helps decode and allocate conversations to segments. In addition to linguistics, predictive Bayesian modelling is applied to render allocation of conversations more deterministic. Such algorithms require sophisticated data science techniques involving expert analyst involvement to create relevant routines on a sample of conversations and then apply these as automated machine learning across all conversations to create real-time applications. The algorithm is predictive of actual consumer behaviour: brand sentiments, computed from the algorithm, have been found to be over 95% accurate in predicting actual user ratings on Yelp.
The ability to replicate strategic survey segments in social media data will help marketers on three fronts: developing social channel, content and targeting strategies to reveal personalisation opportunities; uncovering broader innovation/renovation and co-branding/sponsorship opportunities; and monitoring brand resonance with segment on a continual basis for sensing and responding to issues in real time.
Apart from looking at a mix of social media channels for your segment, which is easy to provide, increasing the engagement of your segment by social media channel is critical. For example, the Contentment segment, which has a very low profile in social media, has a significant proportion of its conversations (50%) as active involvement, i.e. as original tweets or retweets with content. People do express their feelings. However, this does not encourage much interaction, given that this segment need is basically of an
Sunando Das is head of marketing analytics and data science at TNS UK, bringing 18 years of experience across data science, brand strategy and digital insights. His recent work on artificial intelligence algorithms won the Research-Live 2015 best innovation of the year award.