Editor's note: Amy Boren is president of DecisionPoint Marketing and Research, Canton, Ohio.
When I began writing this article I was looking for a simple definition of an algorithm, as it relates to marketing research. To quote from Merriam Webster’s Collegiate Dictionary’s definition, an algorithm is “a step-by-step procedure for solving a problem or accomplishing some end.” In our industry this is leveraged for improved analysis to generate more efficient and effective marketing communications and better optimization.
My experience with algorithms in marketing research usually begins with a product or brand manager requesting a marketing segmentation study, which results in a quantitative methodology study. This leads to findings being analyzed by behavioral variables, grouped into buckets and named. Typically, a follow-up quantitative study is conducted to test the efficacy of the segmentation strategy. With that, a corporate brand algorithm is born. This is the foundation on which the brand story builds over time and allows market research to explain and understand marketing directly to each segment.
When moving into the qualitative phase of research, a product or brand manager extracts the key attributes of the segments and creates a screener for which qualitative recruiters set out to find these characteristics and bring them to life in a focus group facility.
Let me be honest: There is not a project manager in this country that is not intimidated by the task of replicating what the statisticians found quantitatively and relating it back to the general public, within their market, but we do it every day. For example, I work with crisis recruiting for low-incidence studies and I had a client ask me to find males in Chicago who preferred premium plain chocolate ice cream (my client claimed 10 percent incidence). While this sounds simple enough, with all the other ice cream options out there we found two out of 680 people interviewed who qualified and were able to participate in the focus group.
So what about the other 678 people who, in essence, wasted their time filling out the 10-minute survey? I want share with you the potential consequences of this respondent churn in marketing research and what it is going to cost us as an industry in the long-term.
Keeping respondents engaged
Our challenge as an industry is keeping respondents engaged and willing to throw their hat in the ring for a possibility of qualifying for a study. If algorithms are used to profile the coveted respondents, we may see some long-term effects on the general public’s willingness to participate. But in order for marketing research to stay relevant in an organization, it is clear we need to use algorithms to understand the voice of the customer.
In a recent article, “How the marketer/researcher overlap is redefining industry functions and expectations,” Matt Valle of Marketing Insights lists these high profile books/movies/opportunities that brought marketing research mainstream. “Marketing research (inclusive of data, insights and analytics) has come a long way since then,” he writes. “What was once a lower profile function has been vaulted into the public consciousness. Take, for example, Moneyball, the best-selling book and movie about replacing faulty perception with analytics. And there’s Nate Silver, the superstar meta-analytical forecaster, or there’s the emergence of Fortune 500 CEOs with market research backgrounds, most recently Mark Fields at Ford.”
The convergence of new data, better analytics and technology have, in some key areas for marketers, reduced the reliance on intuition and put decision-making in the hands of researchers and algorithms. To facilitate recruiting using an algorithm there is a process, which typically involves administering the screener, augmenting the recruit by taking the responses and running them through an additional Excel tool. These tools are generally pre-programmed with macros that categorize the respondent into dead categories as well as live, with the study only using the respondents that are needed in each live category. Until you have screened them completely you don’t know if they qualify or for which group – sometimes screeners take 15 to 20 minutes to complete.
We all know the “professionals” who will be willing to change wigs, names or underwear if the opportunity presents itself. But what about the regular consumer we want in our focus groups? As researchers, we need to study the long-term cost associated with algorithms in qualitative recruiting. Each moderator and consultant should at least consider asking the recruiter how many overall respondents were screened to net their outcome. At the end of the day our relationship with the public matters in all facets of research.