Seeking the full picture

Editor's note: Based in Vienna, Christian Waldheim is head of GfK’s global social media intelligence center. Based in Boston, Natasha Stevens is a senior practice leader in GfK’s digital market intelligence group in North America.

Today’s social media realm is not for fainthearted marketers. Thanks to social platforms, digital content has already become more participatory – user-generated, highly opinionated – and largely beyond marketers’ control.

Mobile devices are accelerating all of these trends; this is the essence of the new social media experience. Consumers compare prices on shopper discussion sites while standing in store aisles. Through reviews and comments, quickly accessed on smartphones, they share good and bad experiences with brands. “Must-have” products are posted via Pinterest on a host of social platforms.

In fact, social media occupies by far the largest proportion – 29 percent – of mobile online time among U.S. consumers, according to GfK’s MultiMedia Mentor. The runner-up, e-mail, is lagging at 15 percent.

And social media is at the heart of digital “earned media,” in which communications come not through advertising (“paid media”) but the passion and knowledge of journalists, bloggers and consumers – all sharing their perceptions and experiences for free. This leads to dialogue and influence among those who would have never “met” in our offline world. In recent GfK research, earned media had the highest levels of trust among consumers, and therefore (one can argue) the greatest value to brands.

With social media contributing to word-of-mouth around brands with increasing breadth and subtlety, it becomes more and more critical that our social media insight toolbox include the capacity for deeper and highly accurate analysis of social media. We are faced with the classic three-way trade-off among speed/scale, cost and accuracy – in which, typically, you can have two but not all three.

We use machine-based approaches to do social media analysis when speed/scale and cost efficiency are more important. But in a growing number of cases, it is critical to opt for the greatest possible accuracy. Automated platforms, for example, can struggle to address important interpretive challenges, such as: complex expressions with embedded irony, sarcasm and slang; context-dependent sentiments; complex languages; and geographic spillover and multilingual content.

So how can we arrive at a more meaningful approach to social media insights, something that is both scalable and accurate? How can we develop deeper insights from social media to drive market success? We propose a new approach that combines the richness of big data with the subject matter expertise of analysts who can differentiate true, emerging trends from ephemeral ones. This approach, which we call social media insight (SMI), is embedded in decision-making and marketing action at the macro level yet grounded in rigor and accuracy in its development of insights.

The foundation of SMI is an approach that seeks to leverage crowdsourcing technology to produce insights about social media content that are accurate and nuanced yet cost-effective. The SMI process begins with the execution of complex queries of the social media space, customized by channel, to compile a sample of relevant social media content.

We also recruit specially selected and trained local scouts, steeped in the language and culture of the source material. Depending on the assignment, scouts can be sourced from special communities, like medical students or gamers. Scouts review and code these conversations on many dimensions that require a subtle understanding of the content, context and flow of discussion. KPIs can include emotion/sentiment, client strategy-based themes, imagery or attributes and context-based indicators of whether the poster is a consumer or a professional. Used in concert with machine-based systems, this approach aims to allow marketers and researchers to “go deep” into social media to develop useful insights.

We have also found that converting social media analysis into intelligence really starts when social media data is linked to other data sources at scale, such as panel data, syndicated studies and tracking or large-scale segmentation datasets. For example, we have had success applying SMI in the GfK Media Efficiency Panel, an end-to-end panel that passively measures household media exposure, Internet traffic and product consumption. In these environments, one actually begins to learn about cause-and-effect relationships between social media and buying behavior in a broad array of categories.

Applied to the real world

So what does this approach look like when applied in the real world? In one recent study, we combined social media analysis with the Media Efficiency Panel described above to understand cosmetics brands. This study began with a standard assay of social media content about the four leading brands in this space: L’Oréal, Manhattan, essence and Maybelline. The analysis pulled down the URLs and content for over 100,000 posts that mentioned these brands for the period of study. These URLs were then matched to the Internet traffic of panelists to reveal who was exposed to this content and observe the interaction with media consumption and the impact on purchase behavior.

This method yielded “reader”-based estimates of the reach of this content of about 14 million consumers (~140 per post on average). Rather than summing the followers of the posters to get reach, as is traditionally done, the SMI approach gave an independent estimate of what might be considered the “actual” reach. Moreover, because we track consumption in the panel, we were able to estimate that exposure to social media content about a brand is associated with 20 percent more spend on average for those brands.

Finally, scout-based coding can uncover the differential purchase effects of sentiment, theme and even differences between consumer-generated versus professional sourced postings for each brand in the analysis. For example, we observed the effects of elevated consumer-generated content activity flowing from one brand’s long-term presence on Facebook.

In another recent study, conducted in China, SMI was able to better understand the impact of social media content on smartphone brands. Among the key findings of this study was an interesting insight about the nature of social media content as a function of the source. While microblogs like Weibo (the Chinese equivalent of Twitter) produce a lot of consumer-generated volume, the vast majority of social volume contains no emotional content. Conversely, comments about smartphones posted on streaming video sites are less likely to be consumer-generated but contain much more emotional content. So for smartphones in China, it appears that strategies designed to stimulate video content about a brand may be more effective than trying to increase activity on Weibo. Interestingly, this pattern is not unique to China; we see similar effects in western markets as well.

When we examined the main theme of social comments in China for each of the brands in this study – HTC, Apple, Huawei and Samsung – we found that the focus of discussion about Apple and HTC is more about the brand and advertising and the general tone of that discussion is more positive. Discussion about Samsung and Huawei is more likely to be about their devices and can be more critical in tone. For example, this should give those in charge of the HTC domestic brand some encouragement that their advertising and brand messages are resonating in a manner similar to the revered Apple brand.

Finally, in a study of automotive brands in Germany, SMI was able to track the almost viral nature of social media discussion about the Audi brand. In a country where there are three very strong domestic luxury brands – Audi, BMW and Mercedes-Benz – all three have a similar overall share of voice. But the mixture of content for Audi “challenger” brand leans much more to mass social sites like Facebook, while Mercedes-Benz is stronger on more narrow channels like forums and similar Web sites. So when we look at the source of Audi discussion, it is much more likely to be consumer-generated than the other brands in the market, even more than for another brand whose strategy is specifically targeted at driving earned media. Clearly, Audi is making very effective use of social media to drive its brand.

Full alignment

This approach brings the ability to develop insights about social media into full alignment with abilities in traditional media like TV and print. The lessons for marketers and researchers are clear. Brands face a new world of communications and relationship-building in our digital world and social media, abetted by mobile technology, is playing a major role in this user revolution. In order to maximize the value of social media, we must be able to go deep in our analysis, on a global scale, to produce insights that will drive brand success.