The Philosophical Data Scientist: SapientNitro's Stewart Pratt

Q&A: Exec Recognizes the Limits Of Big Data

By Published on .

Stewart Pratt, SapientNitro
Stewart Pratt, SapientNitro Credit: SapientNitro
The hype around data ain't all it's cracked up to be, and believe it or not, that's something that can be heard often from people who work in data-related fields.

Stewart Pratt, director of data and analytics at SapientNitro is one. Steeped in philosophy and economics, Mr. Pratt has unique theories on the world and data's role in it, combining a dedication to humility with a practical recognition of what data can -- and cannot -- do.

"When I speak about the humility of the modern data scientist, I'm referring to receptivity to the limits and role of big data," he said. "Big data can help us identify correlations we may have otherwise missed, but it isn't well-suited for helping us to understand causality or meaning."

Mr. Pratt joined SapientNitro -- a data-focused brand agency -- in 2011. In his data-analytics role, he thinks a lot about emerging technologies, new sources of data and how they can be integrated with traditional ones for clients such as Vail and Jeep.

Among agencies, Sapient is not a bad place for a data enthusiast to hang his hat. The firm acquired marketing-mix modeling firm mPhasize in January, following its purchase of boutique shop Iota Partners in 2012. Iota develops projects that produce new forms of consumer-behavior data at a level that can be mined for insights.

It's not everyday one encounters a data analyst with a philosophy background, but Mr. Pratt majored in that subject at Iowa State.

"While some may find it a bit eccentric, I find that most data scientists tend to be a curious group," he said. "They examine challenges from multiple angles. In addition to statistical models, I'm also intrigued by bigger-picture issues -- for example, how the deluge of data not only impacts how brands gain insights into consumers, but also how it shapes the way people relate to each other and to the world around them."

He steered toward the slightly more practical when he earned a master's in economics at University of Iowa, and later an MBA from the Kellogg School of Management. Mr. Pratt found work in analytics for banks, including HSBC.

"In the '90s, institutions like HSBC kept all the best data and were doing some of the most interesting work on consumer modeling," he said.

Ad Age: A lot of Sapient's clients have their data stored in enterprise data warehouses, where it is difficult for you to mine from them. What are some challenges of dealing with legacy client data?

Mr. Pratt: True, access to proprietary data "behind the firewall" can be a challenge for us agency folks, but interestingly, we often find that a client's organizational processes and methods create greater obstacles. The relational databases many clients depend upon were developed and optimized to maintain high availability. They're built on costly hardware and designed to minimize "failures." Unfortunately, legacy systems often lack the flexibility and scalability required to handle the massive amount of unstructured data that companies are now beginning to collect and to mine.

And while it's costly to maintain these legacy systems, it's the human capital already invested in supporting these platforms that creates the biggest hurdle. A transformation in training and hiring practices is required to truly reap the rewards of big data, as well as a fundamental mind-shift in how companies think about their data and data analysts. Because data has historically been so costly for large companies to collect, maintain and process, they tend to limit the amount of information captured to the bare necessities and adopt analytics techniques designed for understanding small data samples. This is where we find most companies today. But these approaches are generally not well-suited or scalable to a world of cheap, abundant data.

Ad Age: You have a philosophy background. What sorts of connections do you see between philosophic concepts and what works for data scientists and analysts?

Mr. Pratt: It's funny to hear all the misguided perceptions that outsiders have of a "typical data scientist." We come in all shapes and sizes, we really do! Kidding aside, the modern data scientist does tend to share common traits [with philosophers]. One of the most critical being a passion for finding meaning amidst chaos ... an insatiable intellectual curiosity. In essence, they share a love of wisdom, which is, not coincidentally, the literal translation of the Greek word "philosophia."

When I speak about the "humility" of the modern data scientist, I'm referring to receptivity to the limits and role of big data. Big data can help us identify correlations we may have otherwise missed, but it isn't well-suited for helping us to understand causality or "meaning." Humility is a theme you'll start to hear more about. … Just read Kate Crawford's commentary in Harvard Business Review or David Brooks's column, "What Data Can't Do," for The New York Times. Big-data techniques can help us find the signal within the noise, but we still need a human hand to interpret and act upon that signal.

The most common misperception about big data: That it's a science. Extracting meaning from big data is equal parts art and science.

Ad Age: Besides philosophy, what types of educational fields of study and professional backgrounds help develop the best data people?

Mr. Pratt: Many marketers misconstrue data science as the intersection between computer science and statistics. That's part truth. The best data scientists, in my opinion, tend to originate from less-obvious fields of study. Physics and economics, to name two, are critical sources of top talent. Anthony Goldbloom (Kaggle), Hal Varian (Google), James Kobielus (IBM), Nate Silver (New York Times) or Steven Levitt ("Freakanomics") -- among others -- all began as classically trained economists. What's common among these leading minds in data science is the ability to reframe complex problems in creative, simple and understandable ways. Data scientists are the new storytellers for the digital age.

Ad Age: What do you wish marketers would understand about what data scientists do?

Mr. Pratt: That data scientists deliver our best results when we approach our work with clear business objectives in mind but that those goals, ideally, are owned by the marketer.

Ad Age: What's the biggest problem with data science people as they navigate the world of marketing?

Mr. Pratt: Data scientists need to think more like marketers. How? First, keep an eye on improving the drivers of marketing return. Not all relationships we find in the data are actionable or meaningful to improved marketing performance. We data scientists need to be mindful that "interesting" is only truly "interesting" to a CMO if it generates results. Second, focus on storytelling. Great storytelling means delivering the message without losing the audience in the weeds. More often than not, when the message isn't well-received, it's because one of these two principles is being violated.

Ad Age: What's the coolest or strangest type of data set you've ever worked with and why?

Mr. Pratt: At SapientNitro I once worked with a brand to understand how in-console [Xbox, PS2, Wii] advertising placements impact gamers' views not only of the brand, but also of the experience itself. It was astonishing to see gamers respond so positively to advertising when it made the experience feel "more authentic." It's one of many examples I've seen at SapientNitro that demonstrates how dramatically consumer expectations are changing -- in terms of how, and where, they interact with brands. As people's digital and physical worlds blur, their expectations are converging too.

In this article:
Most Popular