Production Recovery Planning in the Time of Disruptive Pandemic: Using Bayesian Thinking to Augment Scenario Analysis

In early 2020 when the extent of the pandemic accelerated in the US, the automotive industry shut down, and the previously anticipated automobile production schedule went to zero with no precise way of knowing when it would recover. Normally, the monthly automobile production forecast that we subscribe to is fairly accurate when the market is in equilibrium, but we know from experience that its accuracy suffers when the market starts to gyrate. The pandemic, of course, imposed a huge gyration on the market. Third party groups supplied us with scenarios that could unfold and rationales for why the scenarios could unfold the way they described, but they did not supply us with any level of degree of belief about which scenarios were more plausible than others. As we moved through the month of April, we needed to know how to reset our planning on production levels, but we had no clear way of knowing which scenario would play out. In this discussion, I share the approach I developed based on Bayesian reasoning to help us classify which scenarios were most probably emerging and to develop a product demand forecast that we could use to plan our production at least through the end of our fiscal year. While this approach is described within the context of the automotive supply industry, it should be applicable to any situation within a planned production environment.