Steve has recently conducted webinars for Gurobi about best practices in applied optimization. Following are excerpts from the Q&A sessions.
Question: How do you use optimization project valuation frameworks and practices effectively for strategic planning?
Steve: As you suggest, it isn’t easy to measure impact in strategic planning. One approach is to build a tool that outperforms traditional spreadsheets by better analyzing alternatives through a deeper, more rigorous simulation. At this stage, lean more on analytics and data science, and hold optimization in abeyance. Then, once your tool is established as the arbiter of value, you can suggest evaluating the results of stretching or breaking certain business rules and introducing optimization to improve the options.
Question: What are the best practices to help educate the black box sponsors to better help customers?
Steve: It can be a bit pejorative, the term “black box.” In this webinar in which we are emphasizing scaled-up implementation, I will say that “getting rid of the black box” is more about the user experience than the creator experience. For a model that is making recommendations, let the end user see why something is recommended, the alternatives that were considered, and how they were scored. That is part of the art of developing practical models and getting them into production.
Question: How do you define “spare cycles” in your discussion of fast recommendations?
Steve: I’m referring to a continuous improvement of solution. In certain optimization applications, the end user wants fast answers—they can be good, reasonable answers, and they don’t have to be optimal. As modelers and developers, we make sure that after presenting the first, “decent” answers, there are “spare cycles” in the background calculating better answers and then threading those answers into the recommendation stack. In other words, there are business cases where it’s not advisable to run the solution to completion to make the best recommendation if that is going to take too long for the end user.
Question: What are your thoughts on open source?
Steve: Professionally, we use a lot of open source software. Sometimes it is not only good enough, it is state‑of‑the‑art. That said, have you heard the jokey question when evaluating specific products: “Is it free as in ‘free beer’ or is it free as in ‘free puppies’”? The total cost of ownership of the software must be evaluated. [For reference, see this representative article.]
With regard to commercial software, we are enthusiastic users of Gurobi. We can send a redacted version of a model to Gurobi and get superb expert support. In the mathematical optimization space, there are only a few world‑class products, and I would be very careful when considering open source and commercial. In many cases, Gurobi is very much worth the price.
Question: Are there any other books for executives on mathematical optimization? Can you tell us a little bit more about your book as well?
Steve: My book, The Optimization Edge (available on Kindle), describes how to practice optimization, what steps to take, and what is difficult about it. Optimization has been subsumed in some sense by the massive interest in AI, which is very encouraging. However, I am less than thrilled that some people equate AI with a specific set of algorithms such as large language models (LLMs) and neural nets. Those are very powerful techniques for certain problems, but they’re not particularly helpful for solving combinatorial optimization problems that Mixed Integer Programming is terrific for. Often, you need a combination of techniques—forecasting, mathematical optimization and more—to solve a business problem.
Currently there is a lot of good writing, though not necessarily in book form, about AI. Everyone I know in the C-suite is being asked what they are doing and thinking about AI. This is our moment! Your group of practitioners might be called Data Science, Operations Research, Analytics, IT—whatever it is, our moment has arrived. Many executives believe this is the great new revolution in business: How do we use smart machines to improve our competitive advantage?
To discuss optimization opportunities with Steve, email us to set up a call.