Talking Decision Analytics Series Part 1: Learning How to Think About Complex Problems

One lesson I came to appreciate while working on a wide range of practical problems is that the exercise of creating computer models teaches us how to *think* about complex problems.

I found that *any* computer model of a complex dynamic problem requires defining five core elements: 1) state variables (all the information we need to make decisions and calculate performance), 2) decision variables, 3) new information arriving from outside the system, 4) the transition function that describes how the system evolves over time, and 5) the performance metrics.

To fill in these details, I need answers to the following three questions:

1) The performance metrics that quantify how well we are doing.

2) The types of decisions we are making (and who makes them).

3) The different sources of uncertainty we have to live with.

Even if your intention is not to create a computer model (or write a single equation), you should always be able to answer these three questions. I continually see books, articles and videos from so-called who talk as if they know something, and yet never address these three questions.

So, start by trying to write down your answers to these questions. This is a good exercise for anyone working at any level of a company, from the C-suite down to driver managers. If this is the last post you read in this series, this advice will help you think more clearly about your company.

In future posts, I will work through each of these questions.  

Warren Powell

Chief Innovation Officer, Optimal Dynamics

Professor Emeritus, Princeton University

Read the introduction: A New Series on Data Analytics for Trucking and Supply Chain Management