Thought Leadership At UC-Davis Graduate School Of Management: Professor David Woodruff On Optimizing Operations

After earning his Ph.D. in industrial engineering and management sciences from Northwestern University in 1990, Professor David Woodruff been at UC Davis Graduate School of Management where he teaches the core Data Analysis for Managers course. It is a course that naturally follows from his research interest in business analytics, operations and management science, planning and scheduling under uncertainty.

Woodruff, who picked up both his M.S. and B.S. in industrial engineering and engineering management from Stanford University, sometimes teaches Managing for Operational Excellence and the Management Science course. Woodruff previously served as associate dean for instructional programs and also as director of concurrent degree programs.

In this wide-ranging conversation, part of the Thought Leadership Series at UC Davis Graduate School of Management, Woodruff looks back on his career and the insights he has developed about optimization. He is interviewed by former Businessweek Executive Editor and Poets&Quants Editor-in-Chief John A. Byrne. The interview has been edited for clarity.

John A. Byrne: David, you are an expert in optimizing operations so that they can perform as best as they possibly can. But my big question for you at first is, why did you become an academic in the first place? 

David Woodruff: Well, I like learning, and if you want to learn something, you teach it. If you really want to learn something, you write a book about it. But teaching works pretty well. And it’s just my job is to learn. So I like having that as my job. When I was an undergraduate, the general academic enterprise was criticized as being something that produces future academics. And it did. I liked that environment. I liked learning. Although I did work at some real jobs for a number of years, I missed the learning and went back and got a PhD. 

Byrne: David, what do you suppose created that thirst for learning?

Woodruff: As an undergraduate student at Stanford, I found out that being an academic meant learning. But like most people, I probably was the most eager learner as a child. Then I found out that, oh, if you’re an academic, you can learn all the time and you have a great environment for learning. 

Byrne: And then the obvious question is why did you devote your life to the field of operations research?

Woodruff: That’s less obvious to me or anyone else. Like many people, I took a few classes, and I said, oh, I like this. Sort of that simple. It was mathematical, but very practical mathematics. At the time, I liked software, and it’s an area where the mathematics and the software come together. 

Byrne: What was some of your earliest research and what did you discover from it? 

Woodruff: Well, I discovered that I like research. As a PhD student, I worked on topics that I don’t work on now. My thesis advisor developed a production control paradigm called CONWIP, and we worked on all aspects of how you would run a manufacturing facility using CONWIP. It’s a production system where you keep a constant amount of work in the system. And so you don’t overload the system with work. You try to keep the right amount of work in the system. It sounds like a good idea. But it’s not so easy for people. I mean, you can imagine that if you’re a production manager and you have things to make, you should start making them. But really, you probably shouldn’t. If there’s already plenty of work in the system, you should wait until there’s not plenty of work in the system and then start working on things. That was the first project. 

Byrne: It’s about maximizing capacity, right? 

Woodruff: Capacity is a property of the production facility. So utilization of capacity would be something you might want to maximize.  If you maximize the use of production, if you get it 100% full, you’ll have queuing. The queues will be out the door. They’ll grow without limit. And the queuing causes all kinds of trouble. The other thing is they’ll be queuing inside the system. So there’ll be piles of work in process all over the place, and it becomes very inflexible. Because if you say, oh, we wish we weren’t making that, well, too bad, it’s half made or half the components are made. So you actually don’t want to use the capacity to 100%. Maybe 80%, 85%, or 90%. If your goal is to make money, as opposed to keeping the capacity 100% utilized, you probably shouldn’t utilize it at 100%. That’s a very tough lesson for people to learn. We weren’t the first ones to discover that. But I think that we produced a system that shows that that makes a lot of sense. 

Byrne: It’s kind of like driving a car, right? If you accelerate too fast, you’re wasting a lot of gas that you otherwise wouldn’t waste because that’s inefficient in terms of gasoline consumption. 

Woodruff: Is it okay if I just totally disagree with you? Sorry, it’s not even close. What it is is, if you want to use a driving metaphor, is if you fill the road up with as many cars as it can handle, then it’ll cause a lot of troubles, cause accidents and cause confusion. In that case, it will actually slow the system down, and that’s true in factories as well. If you get as much work in there as you can possibly put in there, it slows everything down. So although you’re using the road at 100% of its capacity, in some sense you’ve got as many cars on there as can be on there, it’s hurting your throughput, and then if we drop the metaphor and switch back to a real factory, it particularly hurts your throughput of things you want to make. 

Byrne: That metaphor works for me. Now, David, you’ve been involved in a Department of Energy project on sustainable energy. Could you talk a little bit about that? 

Woodruff: I should preface it by saying that I work on general purpose methods, so algorithms, and then we implement them as software, and we produce the software as open source software, so a lot of people use our methods for a lot of different problems. Having said that, we used our method to look at how to optimize energy production. I hate to use the word schedules, but we’ll call them schedules, in the presence of uncertainty due to renewables, so wind and solar were the main ones. The problem that is faced in industry is that they allocate generators a day in advance, and you don’t really know how much wind or how much solar power you’re gonna get a day in advance, so that adds a lot of uncertainty. We worked on how to characterize the uncertainty and how to optimize the allocation of generators in the presence of that uncertainty.

Byrne: What did you learn from it? 

Woodruff: Everything we learned from it was technical and applies to every problem. What we learned that was important is you can take the uncertainty into account. They have very severe limits on how fast they have to solve this. It has to be done in five minutes. That’s what we learned. I could bore you with all the little technical things, but they have nothing to do with wind or sun. They have to do with mathematics and computers and using many processors at once. 

Byrne: David, you have over 9 ,000 citations on Google Scholar. Where do most of them come from?

Woodruff: The most cited works involve CONWIP. That’s an old paper, but it’s the most cited paper in the International Journal of Production Research. So that’s really my thesis advisor’s brainchild, the idea of CONWIP, although I claim important work on that paper.  And then the next one is a paper describing software that allows for a general purpose description of an optimization problem. So anybody who has an optimization problem needs to feed it to the computer somehow. And then I have the statistics work I’ve done that is pretty heavily cited. I work on general purpose methods for optimizing in a sort of mathematical way.

Let me just give a specific. You sell newspapers on a train platform, and you have to decide how many to buy before you go out there, of course. If you buy too few, you’ll run out of papers and you’ll forego profit. And maybe you’ll even lose customers to somebody down the platform. But the main thing is you also have to worry about buying too many because then you’ll have newspapers that are worth nothing, and you’re stuck with inventory. You can’t sell them the next day. So this is a pretty well-known problem, and it kind of characterizes all the things I work on. So if you have this problem and you have general purpose software or methods, then you have to describe the problem, and you have to have a language to tell the computer about the problem. You also have to describe what you know about the data. You may know in advance that the day of the week matters. You have data from the observations you’ve done in the past. You want to give that to the computer and characterize the uncertainty associated with your demand in this simple example. There are other more complicated examples.

Byrne: Makes sense. It’s all about efficiency, really.

Woodruff: I try to avoid that word. t is a good word, but it implies things that don’t strike at the core of what we’re doing.  We want to do the best possible,  and the word efficiency can get people in trouble. If you try to maximize your throughput to capacity, it may minimize your profits.  You may do worse than you think. It’s counterintuitive, but a very important message is that if you load your systems up to the max, you may not make as much money as you would make if you kept them at the sweet spot, at the right amount of utilization, as opposed to 100% utilization. 

Byrne: Your career coincides with the development and the accessibility of computing to everyone, which is kind of interesting in and of itself. 

Woodruff: It is, and the other thing that’s happened, and it’s still happening, is the availability of a lot of processors. So your laptop probably has at least four and maybe eight dual -threaded processors, and to have a 16 -core or a 32 -core desktop is nothing these days. We do projects where we use thousands of cores. And that’s been a big change, to be able to use that computing power to get a solution quickly.

Byrne: When we came of age, computers were behind glass windows in air conditioned rooms and they were massive pieces of iron. I magine that while there had been a lot of development from that period when you started using computers at Stanford and Northwestern, it’s nothing like it was or is today. 

Woodruff: Yes, you’ve heard there’s more fire power in your pocket than there was and that has enabled a lot of people, including myself, to get your work out so that anybody can use these fairly sophisticated algorithms. People have the computers to use them. 

Byrne: You’re heading up a big conference here in July, the International Conference of Stochastic Programming. Tell me about that. 

Woodruff: This was a tri-annual conference. Then COVID happened and it got bumped around a little bit. This is a sort of sub-society of the mathematical programming society. It’s about using computers to make plans. So this conference relates to using computers to make plans or algorithms in the face of uncertain input data. And the conference has people who do mathematics, pure theory, people who do software and applications, and then people in practice who do pure application I’m pretty excited about it. We have a lot of people from Amazon, from the Navy and people from universities all over the world. 

Byrne:  Is there an item on the agenda that most excites you? 

Woodruff: This is my conference, and it’s a week long. So to say there’s one thing that most excites me, I don’t know. To be honest, I’m excited about my own work that’s being presented, so that’s what excites me, the most, I was trying to duck that, trying not to say that, but. 

Byrne: That makes sense. David, you’ve been at this for more than a quarter of a century over that time span. What have you learned? 

Woodruff: The big picture thing I’ve learned comes back to that RPE project. That is, there’s a lot of things you can do. And it amazes me when we go to a conference, practitioners will walk up and say, oh, we’re using your software to do this or that. And we’ve never even talked to them. When I think back to my days as a graduate student, we would write software and no one but me could use it. If you wanted to use it, we’d have to spend a couple of days together. But now it’s made people who I’ve never even heard, met or seen walk up and say, oh, we’re using your software to decide how to manage the solar power on their buildings. If you’re in that position, you have to decide whether to sell that power to the grid, power your building or charge or discharge your batteries. That’s a reasonably complicated problem that you can’t really do heuristically. So you should use software, and some people use ours. It’s very gratifying to know that you discovered something that others find useful.

Byrne: David, it’s been a real pleasure.

THOUGHT LEADERSHIP AT UC DAVIS GRADUATE SCHOOL OF MANAGEMENT

Professor Andrew Hargadon On Innovation

Professor Greta Hsu On Category Constraints

Professor Hemant Bhargava On Technology-Based Business & Markets

Professor Kim Elsbach On Perceptions Of Individuals & Organizations

Professor Paul Griffin On Business Risks

Professor Ayako Yasuda On Impact Investing

Professor Michelle Yetman On Non-Profit Reporting & Governance