Excessive Focus on the Model, Technique or Technology

“A model is a means to an end, not an end itself.” 

“I honestly do not understand all of the math, but I am convinced of the strategic competitive advantage, and significant, tangible, economic value that is created with Yield Management.” – Robert L. Crandall, Chairman, President, CEO of American Airlines circa 1989 (Wharton MBA)

I, for one, have always been enamored with the power and beauty of mathematics. The notation is a language unto itself. Known as the “Queen of the Sciences,” mathematics provides the tools to enable the other sciences, such as physics (which provides the foundation for all of the engineering fields) and economics.

In capitalism, businesses are in business to make money and return that money to shareholders while benefitting society along the way.

At the intersection of mathematics and business, fields like operations research, management science, statistics, and now, analytics and data science are intended to contribute to the betterment of the corporation’s economic and financial performance. The mathematics and models are a means to an end, not an end themselves.

It is not uncommon, especially among recent graduates, to become excessively focused and a bit too enamored with the model and mathematics, the algorithms and technology.

The Pareto principle (80/20 rule) can be of interest and application here, i.e., getting 80% of the benefit for 20% of the effort (or cost). Perfection is the enemy of done! In business, most of the time, there is no need or willingness on the part of management to expend that 80% of the effort to gain the last 20% of the business value. The business needs an answer … and value delivered … now. It doesn’t need to be perfect. It just needs to work and deliver against the economic impact objectives.

The Agile principle of Minimum Viable Product (or Model) is directionally correct and applicable as well. Get to a version that builds, works and generates value ASAP.

In most businesses with which I have worked, the goal was for minimal elapsed time possible to value realization. In fact, in status reports that go to senior management, project update entries must have a business value attached, or they are omitted forthwith. No technical jargon or detail is even allowed. It is implicit and assumed that the correct model form was utilized, tested and validated, as was the business value.

Excessive tweaking, refinements and feature additions or modifications, for little or no measurably incremental gain, are a waste of the company’s time and resources.

I had a team of EMBA students whose final project in a Business Analytics course was focused on improving the accuracy of (binary classification) models to predict mortgage loan defaults operating on a large volume of historical loan performance outcome data (a technique that would have come in handy circa 2008–2009). The students filled 20 PowerPoint slides with mind-bending mathematical models, arcane terminology and symbols, and spent 19 of their 20 allotted presentation minutes talking about all of the different mathematical and statistical models that they had built – Fast Fourier Transforms, Bayesian inference, neural networks, etc. In minute 20, I finally raised my hand and asked, “What was the business outcome result you achieved?” They responded, “Oh, wow, we increased the accuracy of mortgage loan default prediction to over 93%! On their typical loan portfolio, the new and improved model was going to avoid tens of millions of dollars in bad loan default write-offs annually!” To which I responded, “In the future, when presenting, especially to executives, please start with that information.” In journalism, this practice of omitting the most important pieces of information is called “burying the lede.”

The moral of that story should also be applied when presenting to executives inside your company. No one, except perhaps other mathematicians at a conference, cares about all of the technical details. Save that for the appendix. Instead, tell a story of what life was like before and after the model was implemented. Focus on the improved business solution and the incremental business value and economic impact that was achieved in terms of cost, revenue, asset utilization and customer satisfaction – things they understand in terms they use every day. Explain how much time and effort will be saved with a streamlined, automated process. Refer to the controlled experiments that were run to prove out the model’s value, and the testing and validation, with the business domain and finance folks. They’ll want to know that you can “show your work,” but they don’t need to see or hear about all of the math and code.