I always look forward to the September meeting of the Market Research Council. For those who don’t know it, the MRC is a nearly-100-year-old organization, whose early members and honorees included such giants of the industry as Alfred Politz, Arthur Nielsen, Sr., and George Gallup. You have to be elected to membership, so it is an honor to be part of it (check it out). Each year, two people are elected to the Hall of Fame, and the list of honorees reads like a Who’s-Who of the past. The group meets monthly at the Yale Club in NYC to have lunch and a hear from a speaker, so it is both social and professional. Since the group operates on a September to June schedule, the September meeting kicks off a new year, and it is a nice opportunity to meet up with friends I haven’t seen in a few months.
This year’s September meeting was especially memorable for me, as there were two speakers I know well, Bill Pink, the Head of Brand Guidance Analytics for Kantar, and Rob Key, the Founder and CEO of Converseon. In the interest of full disclosure, Bill and I were colleagues and friends at Kantar before I retired in 2015, and I currently do some advising for Rob’s company. I consider them both smart, passionate about the field, and good people. The topic they were to debate, “How Can Brands Best Understand Their Consumers? Social Listening, Traditional Brand Tracking or Both?” is close to my heart (and for cynics out there, no, I had nothing to do with arranging the speakers or topic). I spent a large chunk of my career working on tracking, and my last five years at Kantar working with others on ways to reconceptualize tracking as a predictive system based on integrating multiple kinds of data, including social data, behavioral data, as well as surveys. Those who know me, or have heard me speak, know that I have continued to think about such systems since I left the world of full-time wage slavery three years ago.
The debate itself wasn’t a clash of opposite worldviews; Bill and Rob were in violent agreement on many points. The differences were more in terms of nuance and experience with particular kinds of data and analyses.
Bill spoke first and presented an overview of what Kantar is now calling Brand Guidance (a term I wholly endorse, as I believe I invented it), as opposed to Brand Tracking, to reflect a forward-looking, predictive goal. This system is based on five core principles:
- Speed to insight (i.e. right metrics at the right time)
- Connected data
- Less asking, more observing (i.e. Infusion of Search, Social, Exposure, Receipts, E-commerce
- Migration from descriptive to predictive
- Richer, more granular understanding of brand growth (through deep dives)
Thus, the system uses multiple sources of data to help understand the impact of marketing on short and long-term brand equity and sales. The ongoing survey component is meant to be a “thin core” survey that can be done comfortably via mobile surveys. In a lot of ways, the evolution of this thin core survey over time will prove to be an interesting gauge of how the industry’s use of and comfort with non-survey data grows. Just how thin will this “thin core” become? In my conversation afterward with Bill, we discussed that its evolution will be a function of both psychological factors and analytics – some clients find it difficult to give up reliance on survey data, so they need to be convinced over time that the business benefits from an integrated program will be even greater than what they had before (for a discussion of my own views on mindset challenges within market research, see my article from this past Spring).
On a personal note, it was gratifying to see how the “baby” I helped bring into the world is growing up. I think the work they are doing to look at the evolution of long-term movements in sales and equity, as distinct from short-term movements, will prove especially fruitful and actionable, and reflects the thinking brought together when Kantar combined the different operating companies into one unit.
After Bill spoke, Rob talked about what Converseon is doing with social media analytics, and what it means for how tracking could further evolve. He discussed how the kinds of coding systems usually used limit current social media analytics. Based on Boolean logic, a social media posting is coded as positive negative or neutral, or that the poster is talking about things like “taste” or “quality” based on the presence or absence of specific words or the conjunction of specific words. The quality of results from such systems is surprisingly poor; things like sarcasm and slang notoriously throw the coding off. The approach Rob described that his company is using is based instead on machine learning. Taking a human-coded set of social media posts as a start, a computer “learns” the concepts behind decisions to code individual posts. To put it into terms that my traditionally trained friends would understand, it represents the difference between doing “word coding” of open-end survey responses and doing “thought coding”; but now you’re doing it on steroids, at a scale that humans can’t match.
Besides being more accurate, he argued that now you can reliably code motivations and emotions in social media posts – things like “trust”; almost no one says explicitly on social media that they trust a particular brand. You have to extract the meaning behind what they’re saying to do so. This then enables reliable analyses of higher-level constructs, and Rob showed examples of how they’ve built predictive models of sales and equity using their data. As often happens, Rob was asked by someone about the representativeness of social media data; not everyone posts on social media. While true, the fact that you can build reliable predictive models demonstrates their managerial relevance and value. There is also a developing capability to connect social data to descriptive data like demographics, which would enable social media analyses to focus on particular groups that brands see as their most likely targets.
In many ways then, the answer to my question above about how thin a thin-core survey component can become over time will be driven by the success of systems like that described by Rob; the more value we can derive from social and behavioral data, the less we need survey data. We can focus then on using surveys to provide us insights they are uniquely qualified to provide.
I am sure that there are other companies out there working on solutions like those discussed at this particular MRC meeting. I would welcome hearing from them, so I can write a follow-up article giving a more complete picture of current developments.
Ultimately, as always, it will be clients who drive the direction. So let me close with an excerpt from a recent job posting by Coca-Cola that should make it clear what forward-looking clients are looking for:
“Currently, many of our business decisions are informed by various survey-based tools, which are not connected to each other and don’t reflect the holistic picture that is required to drive high impact business decisions. We acknowledge that to do so, we need to build a new automated Holistic People Understanding platform. This person will shape the development and delivery of the platform by fusing behavioral, social, passive, telemetry, and deep human insights data coming from surveys, and by leveraging technologies, and automation to provide holistic and predictive understanding.”