Eric Schwartz, What Are You Thinking About?



“Big data” has become one of the latest business buzzwords. But Eric Schwartz, professor of marketing, thinks we spend a lot of time collecting data and not enough time making sense of it. Just because we have massive computing power and are drinking from the firehose of real-time data is no reason to ignore decades of sound statistics and economic theory that help us understand what the data truly mean, he says.

Q: What Are You Thinking About?

A: Making sense of big data.

Companies are bringing these two areas together to run experiments. A/B testing helps firms make smarter decisions about what’s best for their customers and what makes the most profit. Which ads work best and where? What’s the most efficient way of acquiring new customers, and what prices should I charge? But in today’s environment we have to not only learn, but earn while we learn. Recently I worked with a company running an online advertising campaign to acquire more customers. We were able to improve how many new customers came in without changing their marketing budget. Instead, I adapted machine learning methods to the marketing, leveraging an area of research on the “multi-armed bandit” problem. We had to learn which ads worked best on which websites in real time over hundreds of millions of site visits. Over the course of a few months we adjusted the ad allocation and improved customer acquisition rates by eight percent versus a standard A/B test. But it’s more than just which ad gets the most clicks, and it’s more than the number of new customers. You want to maximize the long-run value of customers per dollar spent on acquiring them.

Why is this interesting to you?

I have a real taste for this kind of work. I like the theory, I like the practice, and I like the instant application in the real world. I’m always interested in thinking like a scientist and an engineer to solve problems for organizations, as well as asking questions of general interest for research. There’s so much excitement around big data, but sometimes we’re not thinking enough about it. As marketers, it’s our role to make sure we don’t forget there’s centuries’ worth of good economic theory that we should stand on. We need to make sure what we do with machine learning and statistics is causally sound. I think it’s important to bridge these gaps.

What are the implications for industry?

How you run an ad campaign, how much you spend on customer acquisition, and how much you spend to retain them are all age-old marketing problems. But how you adapt to all of this in a digital environment where the data comes at you in real time is the new twist. You can drown yourself trying to look at every single piece of data. So my research stream is focused on making sure we don’t test minutiae. How can we get to what really has an impact on customers and profits? That’s the big challenge for companies today.