The Rise of Machine Learning

• AccuWeather, based in State College, Pennsylvania, is leveraging the proliferation of data to enhance the specificity of its forecasts down to a street address and a one-minute window, extend the length of its long-term prognostications to 45 days and add a social-media feedback mechanism that augments its reports with individuals’ photos of conditions where they live. “These things are valuable to our clients,” CEO Joel Myers says. “And the more valuable our forecasts are, the more they improve our image and the more people realize we are the go-to source for accuracy, precision and detail.”

• Intel has used its Big Data expertise to diversify into healthcare over the last couple of years with wearable monitors, ranging from 24/7 tracking of Parkinson’s patients so they can get improved treatment to the introduction of smart headphones with sensors that can measure heart rates and which carry the lifestyle brand of rapper Curtis “50 Cent” Jackson.

• Philips North America is analyzing data from its medical monitors to predict which users will end up in a hospital in the next 30 days, helping acute-care providers intervene to alleviate chronic conditions at home and pre-empting the need for expensive readmissions to a hospital.

• AMN Healthcare Services is crunching data with predictive analytics to extend scheduling of nurses and its other contracted staffers as long as four months, with 95% reliability that it will work out satisfactorily for each staffer, by using patterns such as the absenteeism and overtime histories of individual nurses. That compares with the historical practice of being able to schedule out only two or three weeks—and on a hit-or-miss basis, at that.

“It’s a big dissatisfier if nurses can’t plan in advance, and nurses are about one-quarter of a hospital’s budget,” explains Susan Salka, CEO of San Diego-based AMN. “With nurse attrition at historically high levels right now, anything you can do to keep them happy is a financial benefit to you, because they’re more likely to stay.”

Making Big Numbers Work
Effectively leveraging Big Data and machine learning requires more than a CEO’s determination. Here are some other considerations:

Now is the time
It’s feasible now to do machine learning at scale and in real time, because of how capabilities have improved and costs have come down. Yet CEOs will have hesitations. “They’ll have difficulty maintaining trust and balance with these learning systems,” says J. Lance Reese, COO of Limu, a Lake Mary, Florida-based maker of energy drinks and other beverages. To overcome this problem, he advises CEOs to “have more direct contact with technologists” and others who direct machine-learning systems.

Boards are expecting action
Directors are the primary influencer—above CEOs or CIOs—of Big Data adoption strategies in 53% of the companies surveyed by Accenture and GE.

Where companies are striving to make an impact with Big Data within a year or two, board members are driving much of the urgency. That’s one good reason to “make data and analytics a larger part of every board meeting,” John Kelly, leader of Berkeley Research Group’s predictive analytics practice, advises CEOs.

Selectivity helps at the start
The best way to begin leveraging Big Data is to “have a hypothesis and act on it and test the data to see if you can harness insights” rather than “put all the data together and see what story it tells you,” says Anshu Prasad of A.T. Kearney. That can be difficult for CEOs to do because “all the function leaders are saying, ‘I have to do something with Big Data and analytics,’ so 1,000 tulips are blooming.”

Start with experimentation
Probe the capabilities of machine learning in a limited way at first. “Get someone to play around with it,” advises Altair CEO Jim Scapa. “Get a smart kid.” That’s why one of Altair’s first forays into machine learning is a project at Scapa’s alma mater, Columbia University.