Predictive analytics for non-analysts: interview with Randy Bradley

As most practitioners can tell you, there is a treasure trove of historical data available to manage today's supply chains—too much data, in fact. The problem isn't so much obtaining the information; there are many good tools around for that. The real challenge is making good decisions from all of it. That's where having a "prescribed" approach can help.

Dr. Randy Bradley has spent a good portion of his career looking at analytics data and figuring out the best ways to analyze it for beneficial results. You might say he is an analytics analyst. Dr. Bradley is an assistant professor of information systems and supply chain management in the Haslam College of Business at the University of Tennessee. He holds a Ph.D. in management of information technology and innovation, an M.S. in management information systems, and a B.S. in computer engineering, all from Auburn University.

DC Velocity Group Editorial Director Mitch Mac Donald spoke recently with Dr. Bradley on the value of predictive analytics for digital supply chains. You can watch the full interview or read an edited version of their conversation below.

Q: You did a presentation at MHI's fall conference titled "Predictive Analytics for Non-Analysts." Tell us a little about that.

A: For the non-analyst, the hard work of analytics is really the strategic approach. So I talked a bit about three key components of that—data streams, questions, and strategy—and why organizations need to focus on them.

When we look at the emergence of technology in the supply chain, we have more data coming at us than we know what to do with—and it's coming at us from a multitude of different angles. Organizations need to have a cohesive process for how they're going to manage that flow of data. We already know the issue with silos. Now, you lay on top of that more data sources from other silos. That is why there is a need to focus on the data streams first. How you are going to manage the capturing, the processing, and the structuring of that data?

Then, the next thing is the questions. Oftentimes, what I find when I work with executives is that they ask one question, but really, they want an answer to another question. The reason why questions are paramount is that questions drive the mechanism to get to the solution. In other words, analytics is not about which algorithm I'm going to use or which solution I'm going to apply. It is about what question am I trying to answer. The question drives the approach or the technique.

Then, last is strategy. You'd be amazed that approximately 80 percent of the organizations we work with don't have an analytics strategy. So, we are shooting for something, and yet we have no guided direction with respect to that.

Q: So, if you're a non-analyst, you should focus not on the algorithms and the codes and the databases and where things link and how things get shared, but rather on the questions you have and how analytics can deliver answers to make better business decisions?

A: Absolutely, because people are enamored with predictive analytics, but predictive analytics essentially tells you what is likely to happen. We don't know that it will. It is just what we believe based on the historical data we have. But the better question is, what steps am I going to take in the event that it does happen or in the event that it doesn't happen? So, we are trying to get from just predictive analytics to prescriptive analytics, where we have a prescribed approach to a particular incident or outcome.

Q: Predictive analytics is very hot right now. Why is that?

A: I think it is the nomenclature itself. When we hear the term "predictive," we think of it as this perfect picture. We think we know exactly what's going to take place, or that we'll have what I like to describe as a heads-up view in an automobile: You know the direction you're going, you know how fast you're going, and you know when your next turn is. People think that is predictive analytics, but it is not. Predictive analytics is the rearview mirror. The best vision you have of what's in front of you is really what's behind you. Everything else is hazy. So, you're going to use your historical data to try to anticipate what is likely to take place.

Q: The industry is full of very bright people who are passionate about what they do and who are not resistant to change. They see this coming and recognize it is important, but they don't know where to begin. What is step one for these folks?

A: For me, step one is strategy. The reason I say this is that the data streams and the questions are going to be contingent upon your strategic approach and the strategic imperative you place around analytics in the organization. We say we want to be data-driven organizations, but that can mean 10 different things. So, the question is, what does it mean and what should it mean for my organization? Once we put a stake in the ground, this is our analytics strategy.

That is not to be confused with a big-data strategy. You already know you live in a world with voluminous data. You don't need a strategy around that. That should be embedded in your IT strategy, which should be coupled to your business strategy. But your approach to analytics in driving decisions should be to view it as decision analytics. Analytics gives you the vision to analyze the types of decisions you've made in the past and whether or not they've been fruitful.

Q: You just made a point that I think is fascinating—to make sure your IT strategy is not in a silo serving itself but reflects the goals of the company. Let's shift a bit to the broader topic of technology. I know it is a passion of yours. A lot has happened in the last 10 years, and a lot has changed. What is your short list of some of the most important or disruptive technologies that have emerged in the supply chain in the past decade?

A: I think that we're now embracing artificial intelligence to the degree that we really should. And a lot of times, we use the terms "artificial intelligence" and "machine learning" interchangeably, but the reality is, machine learning is a subset of artificial intelligence, or AI. We used to talk about "hard AI" and "soft AI." Machine learning is that soft AI, where the programs learn to do things that you never instructed them to do.

I think we now have the computational power, we have the storage capabilities, and we have more data coming, so we can truly feed those appliances. It's what's hot now, and I think that is what is going to carry us for the next 10 years. I think we are going to get true insights because we are now finally able to harness the power of something that was created back in the '60s.

Q: What are some other technologies that you think are going to go from promise to application in the next three to five years?

A: I think we will see blockchain. I really do. A lot of people are interested in blockchain. But when I say we're going to see promise with respect to that, I think we are going to truly realize what it can and can't do. I think we are also going to realize where we should and where we shouldn't use it. To me, once we get that level of clarity, that is when I think solutions start to become more tangible.