to create models in one platform and move them to another to be used. In business today, models become obsolete very quickly as tastes, demographics, products, and other fac- tors change rapidly. This new process creates new models during every time it is used so they never go stale. Moreover, with this new process, the user doesn't have to be an expert because the expertise is in the computer. HCOB: Can you tell us how this process is used to provide real business value? Hale : Our first implementation was for merchandising analysts who thought they were just pulling reports when they were actually using machine learning. I met with the VP of Marketing for a large retailer to discuss his most important problems. His most pressing issue was that he couldn't keep up with his requirements to provide targets for promotions being offered by his vendors. He had large amounts of data including massive transaction data and loyalty information on tens of thousands of customers, but no way to create predictive models to identify the correct targets in a timely fashion. There had been lots of attempts to do hands-free predictive model creation by automating the processes around tradi- tional techniques, but I wasn't aware of any that had been widely deployed. I decided that to meet the needs of predic- tive modeling in the world of machine learning and big data we would have to use something other than the traditional techniques. I took the opposite tack of most research and started working with simplifying the analysis and relying on the massive amount of data available to improve accuracy. It not only worked, but it offered results similar to those provided by more traditional techniques. The relatively simple techniques used produced significantly less processor
overhead enabling model development and deployment to be accomplished within the data warehouse, and not in a special-purpose processor. The patented process has been deployed around the world and used for solutions other than promotion targeting. Several of the world's largest mass merchandisers use it for replenishment prediction and have enjoyed reduced out-of- stocks and increased turn. It's also been used in healthcare applications, and its automatic operation has also given it a home in e-commerce. HCOB: It’s fairly unusual, isn’t it, for a businessman to create what is essentially a very technical, engineer- ing-based process patent? Hale : Well, yes, I suppose it is. I held titles of Director and VP in sales and marketing and yet I was awarded a patent for a predictive modeling technology, which one would think would come from a research or engineering back- ground. That's because I had training in general business and predictive analytics at Auburn and a developed a deep understanding of marketing needs from my business career. That’s one of our goals in the Harbert Business Analytics curriculum — to produce graduates with a strong under- standing of both business analytics and business processes. Graduates with an understanding of both worlds have a distinct advantage in today’s increasingly technical business environment. HCOB: What advice can you give to today’s business students looking to leverage the experience and entrepre- neurial spirit you’ve embodied throughout your career in their own professional development?
Hale : Entrepreneurs are driven creators who take chances and try new things in virtually everything they do. Entrepreneurs exist in startups and in large organizations. They’re the people who aren’t afraid to try something new and different. And there's no better place to start than in the world of business analytics where big data and advanced analytical tech- niques are opening new possibilities every day. That’s why Melinda and I are so committed to the work David Paradice and his team are doing at Harbert in Business Analytics.
The Richard and Melinda Hale Endowed Fund for Excellence for Business Analytics was founded to help David Paradice, Harbert College Eminent Scholar in Business Analytics, match Harbert ’ s Business Analytics curriculum with the emerging needs of today ’ s data-rich business environment.
47
Powered by FlippingBook