This article was first published by Oliver Wyman as part of its report in April 2020.
Companies have traditionally relied upon consumer surveys, focus groups, and research reports to figure out what people think of their products or services. But these approaches have several shortfalls. Sample sizes are limited and subject to bias. The studies take time to organize, and the results quickly become dated. Moreover, what people say often differs from what they do, as when they complain about discount airlines but then use them all the same.
Social listening provides an alternative: It enables companies to tap into much richer consumer insights that are generated in real time. Until now, social listening has mostly been limited to public-opinion monitoring. For example, it is used to count the number of times a brand is mentioned 鈥 鈥渂uzz鈥 鈥 and whether the content is positive or negative 鈥 鈥渟entiment鈥. But this monitoring tends to yield only standardized or aggregated metrics, which seldom lead to actionable business decisions. Recently, advances in machine learning have made smart analysis of natural-language content possible, as well as the monitoring of pictures and videos. The new techniques have been very effective in enabling companies to investigate a wider range of consumers鈥 feelings and follow different aspects of their lives. We have seen how their preferences can be mapped and updated immediately, as can the ways in which customers connect with and influence each other.
Machine learning can also overcome the problem of 鈥渦nclean鈥 data. Companies often buy social media data from a vendor, but many of the source posts do not actually come from consumers; instead, they have been written by the brand鈥檚 marketing agency, e-commerce sellers, or robots. We discovered one case in which more than nine out of 10 posts provided by a data vendor had not actually been written by consumers. Such posts are typically overly positive and give a false impression of a brand鈥檚 health. Machine learning can analyze social-media accounts and their content in order to filter out such fake posts.
We are only beginning to scratch the surface of these advances鈥 potential to rewrite the rules for consumer product companies. Already, social listening is overturning development, marketing, and packaging. One manufacturer doubled its rate of hit products from one in 10 to one in five thanks to better consumer information, halving the cost of developing also-ran products. Various use cases will often be combined. A retail bank launching a greenfield digital unit, for instance, can first use social listening to identify and quantify customer pain points, segment potential customers, and inform product design. After launch, influencers can be sought out and targeted, and the brand perception tracked, in order to optimize products and services continuously.
Social listening techniques have been around for over 10 years, but very few companies are making the most of them. The main barrier has been the quality of data, which are usually too aggregated to provide significant insight into consumers, or else are based on social postings that include large numbers of fakes. Moreover, obtaining clean data that inform business decisions is not easy. It can require natural language recognition in different languages and advanced techniques to analyze social media accounts and their contents automatically and effectively. Soon however, these techniques will be common in consumer-facing industries, and those companies that neglect them will find it hard to catch up. By contrast, product manufacturers that learn how to understand consumers in actionable ways 鈥 and to test consumers鈥 reactions in real time 鈥 will have a huge advantage.
