Professor Thiago Serra is an assistant professor of business analytics at Bucknell University. He completed his doctoral work at Carnegie Mellon University and has spent time working in the transportation and energy industries. He recently invited Nextmv CEO Carolyn Mooney to present to his business management class about decisions as code. We wanted to take the opportunity to turn the tables and pick his brain about the optimization world.
This interview has been edited for length and clarity.
Thanks for taking the time to chat with us today, Professor Serra. Let’s start with the basics. How do you describe optimization?
There is something tricky about using the word optimization in the age of the internet. If you search for it, search engines will often produce results about search engine optimization. So I have joined others in saying mathematical optimization for what we do.
For many people, this means there is a matrix, there are mathematical expressions, etc. You can also frame it another way with more of a software development approach, as you’re doing with Nextmv. But at the end of the day, you’re still searching for a minimum or maximum solution that adheres to a given objective function while accounting for specific business logic.
Why should we care about mathematical optimization?
One thing that I like about mathematical optimization is that it works like magic: you can find a more efficient plan that uses basically the same resources but produces something that is much more valuable for an organization. For example, say you have some orders to deliver. Mathematical optimization allows you to find and choose a shorter route so that you can accomplish the same thing, but faster and using less resources.
This is both a blessing and a curse. When done correctly, optimization can be invisible to the customer: they walk in, get what they need, and leave as if nothing out of the ordinary happened. But when optimization is missing, the customer will notice: prices aren’t quite right, service is too slow, products are not displayed properly or working in the best possible way, and so on.
Taking a step back, how did you get your start in the optimization space?
I learned about optimization through some courses in college and by pursuing an undergraduate research project. Then I went on to work as a software engineer when I was about to graduate and quickly got bored of the types of projects that I came across. I found this consulting company in need of more optimizers and I took a chance to try something more challenging and meaningful. While there, I worked in factory scheduling, routing deliveries, and organizing orders in containers. For the most part, I was still a software engineer. However, I could see a bigger impact in the work that I was part of.
How much has changed in the optimization space since you got started in it to today?
This is a very personal impression, but I believe that people are prioritizing ease of use over performance more and more. When starting my PhD, I was absolutely sure that continuing to use Java would not be acceptable and I pushed myself to learn C++ in depth. When finishing my PhD, I was working more and more with Python even though that meant a tradeoff in performance. It has been a while now since I last used a lower-level programming language for anything.
As you teach classes each semester, how are you seeing the skill set change and develop for the next-generation workforce in this field?
Teaching mostly undergraduate management students at Bucknell, I am observing a greater expectation from the job market that they should be able to develop some proficiency working with data and with a computer programming language like Python. This allows them to be able to analyze data beyond what a spreadsheet is capable of.
We do make sure that our business analytics majors develop such ability, but I am seeing an increased amount of students from other majors coming to our classes for the same skills. Some of which in reaction to the expectations set forth by potential employers.
What’s the biggest hurdle you see for students as they transition from optimization in an academic setting to optimization in industry?
I have an observation that goes a little beyond optimization: my main concern is when the choice of the tool precedes understanding the problem that should be solved. For example, when the first thought of a manager is not about understanding the problem that they need to solve, but instead to figure a way to use optimization, machine learning, or artificial intelligence on it.
I believe that this is less of a problem in the case of optimization because we are in a model-driven subject: we define the model based on domain expertise rather than relying on the data to tell what the model looks like.
However, ignoring relevant relationships between the parts of the business that we want to automate with optimization can render these models useless as well. In other words, my main concern is with students jumping to technical conclusions too fast before understanding the context of the business problem in front of them.
Who’s doing interesting work in the optimization space at the moment?
That is a tricky question to answer because I do not want to commit the injustice of forgetting someone important on the spot. I have seen many colleagues rethinking what we do with optimization in so many ways. For example, by thinking of human welfare and environmental protection as the objective function rather than cost, revenue, and profit.
I am also very excited about the new and unexpected connections between optimization and machine learning, in particular when optimization takes machine learning to new heights, which is the kind of work that I am mostly involved with.
What recent optimization reading or watching recommendations do you have for folks?
This is not exactly recent, but I believe that every applied optimizer should read Red Plenty by Francis Spufford. This book follows some fictional and historical characters during the years in which the Soviet Union tried to invest in what used to be called cybernetics as a means to improve their productivity and keep up with the West.
I have read some excerpts from this book at the beginning, middle, and end of my recent course on optimization for management students. It is very interesting to see how early pioneers of optimization failed to understand the human side of the problem and never got to see their plans implemented at the scale that they hoped. In a sense, this is exactly what we discussed before.
///
To learn more about Professor Serra, visit https://thiagoserra.com/. You can follow him on Twitter at @thserra and watch presentations he hosts about optimization in sports scheduling, political districting, and more.
