Start of Main Content

Capture.JPGBlake McShane is an associate professor of marketing at Kellogg, where he teaches courses in customer analytics, marketing research and data analysis. As a statistical methodologist, McShane has developed statistical models for a variety of fields, including online advertising, neuroscience, paleoclimatology, law and even baseball.  His research primarily focuses on developing new methodologies that accommodate the rich and varied data structures found in business problems.

Read on as Professor McShane discusses his marketing research course, his research on p-values and what he loves most about teaching at Kellogg.

What courses do you teach at Kellogg?

Starting this fall, I will be teaching Customer Analytics (MKTG-953), but for the past six years I have taught Marketing Research and Analytics (MKTG-450). MKTG-450 is a broad survey course that covers all aspects of marketing research and its different methods: qualitative research, causal research, descriptive research, questionnaire design and more.

The focus of the class is an experiential learning project, where a group of five students is paired with a client. Our clients really run the gamut, including the Chicago Bulls, several consumer packaged goods brands, an insurance company and U.S. Bank. (The students are helping U.S. Bank with its mobile application.)

All of these companies come to Kellogg with a marketing research problem, and we have our student groups work on these problems. The course is structured so that students start the quarter figuring out what the objective is and fine-tuning it. They then look at secondary data — what data is already out there that speaks to the problem.

Next, students conduct basic exploratory research and create a quantitative survey to obtain data from a paid panel of respondents. The composition of the panel is tailored to the specific marketing research problem that the students and their client are facing so that they get good, high-quality data. We aim to get 200-300 respondents per group so that our students have a relatively large data set to analyze, one that allows them to offer rigorous recommendations to their clients. The ultimate goal at the end of the course is to produce an actionable recommendation.

What do you hope students take away from your course?

First of all, I hope that students get a satisfied client out of it. I also hope that my students have become sophisticated consumers of research who can understand and critique it; after all, the vast majority of them are never going to be marketing researchers themselves.

Some of being able to critique and consume market research in an intelligent manner involves learning by doing. That’s why we have the experiential aspect of the course; getting your hands dirty and conducting research yourself allows you to understand the choices that are involved and some of the more subjective factors at play.

I view market research as something that’s not only useful for traditional marketing brand manager roles, but also for consultants who also are consumers of marketing research. Thus, I believe this class is useful for a variety of careers, even though it’s a very traditional marketing research class.

In the era of big data, what are your thoughts on balancing quantitative and qualitative data?

The traditional marketing research view of the world (and this probably encapsulates a large fraction of research out there) is that you start with secondary data and qualitative techniques at the beginning of your research to figure out what it is that you don’t know. Then, when you know what it is you don’t know, you use quantitative methods to figure it out.

I think that is the approach still used in a lot of marketing research, but sometimes you can use qualitative techniques at the end instead. Or sometimes they can be helpful in the middle. There’s no one right way to do it; it’s very problem-dependent.

Also, I’d be careful when looking at historical data, whether it be internal customer records, third-party data, census data or data you get from a vendor. It’s important to be sure that you’re not just looking at correlational patterns. We want to make sure that if we’re contemplating business interventions, like a new promotion, that the promotion was really what caused the outcome of interest (e.g., product trial). Oftentimes when we just throw big data into the grinder, we end up with correlations that might be meaningless. A lot of MKTG-450 is about thinking quite critically about these kinds of issues.

What can students learn from your p-value research?

I would say there are two related parts.

Very often, the way people think about a problem is: I have a data set, I run a statistical procedure and then I get a p-value. If that p-value is less than .05, I’m good to go! There are at least two shortcomings to this way of thinking.

One is that, as a recent Kellogg Insight article explains, there’s nothing sacrosanct about .05. You shouldn’t really be making decisions based off of .05. You want to take a broader view and, for example, think about, What was the size of the effect that I measured? Do I think it will be reliable and generalizable to different contexts? Is this intervention cost effective? There are a whole host of considerations you want to consider beyond p-values. That’s one aspect.

The other aspect is that we rarely have only one data set, for example one study or one view into the phenomenon. Typically we have multiple data sets that speak to the same question, maybe from different aspects or in slightly different contexts. Another part of my research involves what to do in this context when we have multiple views of the same problem that are all sort of similar but also somewhat different? Statistically, how do we knit them together in a rigorous and holistic way?

What do you like most about teaching at Kellogg?

In every class that I’ve taught at Kellogg, there’s been a handful of students who teach me something. There’s a collaborative aspect, where students can bring experiences that they’ve had and enrich the whole class with those experiences. Students can add some nuance and texture, especially the Evening & Weekend students who often bring fresh experiences from that day or week at work.

Overall, what I really like about teaching at Kellogg is the quality of the students and the variety of their experiences prior to Kellogg — what they bring to the classroom is tremendously enriching and helps create a collaborative learning environment.

Want to learn more about Kellogg’s data science offerings? Check out our Program on Data Analytics at Kellogg (PDAK)