Data Models are Reflections of People
In January 2024, I began the Accelerated Management Program at the Yale School of Management. I enrolled in the program because I wanted to learn the vocabulary and mechanics of accounting and finance. We studied linear programming, which is a mathematical technique that helps find optimization in the face of multiple constraints. After completing the readings, lectures and assignments, I started to think more deeply about constraints (e.g. time, money, resources) and the thought process (reasoning) behind their inclusion and exclusion in model development. I kept asking, how is it possible that reasonable people can look at the same data and information and come to very different conclusions about what is important (worth including in the model), how much weight constraints should be given, and where to place resources and capital?
I believe there is a multifaceted answer to this question.
I adhere to a very simple adage: if you know the question you are asking, the answer will readily reveal itself. The idea being that it takes much more work and time to gain clarity on the underlying issue and problem. What is really difficult, after you understand the problem, is figuring out what constraints or dynamics are not informative or worth including. The more dynamics (constraints) that are included, the higher likelihood that the proposed solution will be unusable. Why might this be the case? Simply because of the spillover effects that must be accounted for — that is, thinking through how each constraint informs one another and, potentially, the unintended consequences that may arise from their interaction(s). I have to be very clear on why I think certain constraints are worth inclusion and exclusion. This is where and why models are more art than science.
For this reason, I believe mathematical models and their underlying assumptions are simply a reflection of the person constructing the model. The model, if I place the math to the side for a minute, just shows me how someone is thinking about the relationship between individual pieces of the model (e.g. production costs and shipping) and how those pieces inform and build up to a desired outcome.
This raises an interesting question worth considering. Do good decisions lead to good outcomes? Not necessarily. The model is a tool. That is all it is. If you don't know how to use or refine it, then every problem, question, scenario desired outcome looks the same. It is a new take on an old saying, "if all you have is a hammer, everything looks like a nail.
