Mining the Missing Link in Data Analytics: Human Behavior

humanbehaviour
Human behavior still could throw even well-crafted data models out the window

The data-driven enterprise is on the rise, and there are highly sophisticated tools and technologies available to make things happen. But there’s still a big question mark around one of the most vital components of the data-driven world: human behavior. It’s unpredictable, quirky, subject to various motivations and un-motivators. The question is, can enterprise executives really know they understand what’s going on, or about to happen, if they don’t have a firm grasp on this aspect of their customers, employees, or operations?

Building human behavior into analytics data models – or at least accounting for it — was the subject taken up by a panel of corporate data leaders at the recent Cloud Business Summit, held in New York and hosted by Saugatuck Technology, a division of ISG.

Panelists agreed that human behavior is still an unknown that could throw even well-crafted data models out the window. At the same time, data analytics is starting to make some headway in better understanding patterns of human behavior. At the same time, finding or promoting the skills to help in this effort is challenging, and will continue to be.

The panel, led by Saugatuck analyst Jim Hurley, started off with a discussion of Google’s self-driving car, and what the demonstration project has already taught the industry about the possibilities and limitations of data-driven decision making. Mike Lemberger, vice president of global information products for Visa, says while many of these challenges are already addressed in today’s aircraft, automobiles have a way to go. “If you think about the airplane, the pilot takes off, puts it on autopilot, then lands it,” he illustrates. “If you think about the car, the intelligence is in those unique components — starting and stopping, stop sign the red light, the person in the crosswalk — requires the fine-tuning of the end points, and I think that’s where advanced analytics comes in.”

Irving Wladawsky-Berger, visiting lecturer at MIT and former executive vice president of IBM, points out that the “science” in “data science” may actually too limiting. “If you look at natural sciences, it’s simple — you have galaxies and planets and electrodes, they will do what they sort of do, and you can statistically control for what they do,” he said “If we apply that to peoples’ behavior, who the hell knows?”

That’s why analytics-driven enterprises need to extend their competencies well beyond the statistics and number-crunching that are part of data science. With the “gazillions” of data points now being collected from customers, employees and partners, there’s an opportunity for businesspeople, with the assistance of data scientists and analysts, to delve in and better understand patterns of human behavior. “You have all this data, but then you need the really good marketer, the really good business executive to interpret it and then when you interpret it,” he said.

In healthcare as well, there is a need to move beyond the clinical data that is being collected in abundance and develop a more holistic understanding of patients. “Right now, were emerging from the Bronze Age in health care,” remarked Tim Gilchrest, director of eBusiness strategy and innovation for Anthem Health and fellow at Health Innovation Lab at Columbia University, who noted that right now, medical practitioners may have about 10% of the data they need on patients. “We collect telemetry, blood pressure and labs in the healthcare space and we try to put that together and make a guess,” he stated. “What we’re learning is the social data and wearable data, is the stuff that really matters. How far you ran today is much more important than your lab results from six months ago, and much more telling in terms of what will happen to you.”

This is significant, he continued, because the typical approach to medicine has been to attempt to urge people to take certain drugs for their conditions or illnesses. Perhaps, with more holistic data on patients, more successful and less coercive therapies can be applied, through behavioral economics. “What are some of the levers that are present that will help change your behavior in a good way?” he asked. “Maybe you’re a diabetic because you’re lonely, not because you eat too much. Instead of sending a message that says stop eating so much, maybe we could recommend they adopt a pet. We’re moving out of clinical data and moving into social data.” Gilchrest added that within the Columbia University lab, practitioners have been able to diagnose diabetics, via social data, with “80% accuracy.”

Staffing the analytics data-driven organization is another challenge. Organizations need to focus on achieving a balance in the skills and aptitudes essential to a data-driven enterprise. Data scientists are one piece of the equation, but organizations need more than strictly statistical or mathematical capabilities, panelists concurred. “If you’re a businessperson, before you pull the trigger on this being machine learning initiative, you should probably educate yourself,” said Will Klancko, enterprise data management leader for GE Capital.

Most importantly, organizations need to develop analytical cultures that will pervade the way things are done. “Don’t be afraid to fail — open up your mind, and try stuff,” Lemberger urges. “Too many companies think they’re going to get all this data, and that it’s going to be perfect. You have to accept the failures and the data may lead you to the wrong place. Train people to look lots of places, and that it’s okay, because some of the conclusions may be wrong. Data doesn’t lie, but it does provide false positives.”

Experimentation and failure with data needs to be part of the equation since data analytics is still a very young field, Wladawsky-Berger pointed out. “It’s a really new discipline, this stuff is really new. It’s important to start getting some people more comfortable with asking the right questions, and techniques like machine learning or statistics, and how to apply this learning to the business.”