Forecasting, Risk and the Supply Chain

Forecasting changes in the supply chain is typically done through two broad methods: judgmental forecasts and quantitative analysis (per Wheelwright and Winslow).  Judgmental forecasts are based on the opinions and intuition of managers and experts, and are generally developed to account for different cognitive biases. Quantitative analysis uses different techniques (such as time-lapse forecasting) to evaluate past data and predict the near-term future. Both of these methods are reliant on using past performance—whatever trends that have already happened in the supply chain—to determine the future by creating approximations of what is most likely to happen in the next day, week, month or year. Everything that deviates from this projection at any point, or the linear trend over a period of time, is considered an aberration that doesn’t account for what’s most likely to happen. 

While this thinking is rational and very useful in decision-making, disregarding or minimizing the importance of occurrences least likely to happen can be dangerous, due to 1.) the assumptions businesses make about the future based on past performance and 2.) how the faith put into the “hard” numbers produced by quantitative analysis can lead to cognitive biases (optimism, groupthink) at the managerial/ decision-making level.  

These two themes contributed to the 2008 financial crisis, as several major banks based their decision-making on  Value at Risk, or VaR, models. VaR models reduced complicated data down to a single number that captured the maximum amount a bank could lose in a given time period with a high level of confidence (99 percent generally). Lenders based their decisions on the VaR number, which was issued to executives on a daily basis. One of the major flaws of VaR was that it didn’t account for risks that were possible, but that had not happened within the timeframe the model was using to create future predictions. For example, housing prices either rose or remained stable (1960s-2000s) in the timeframe used by several banks in determining VaR. So, because of the time-frame used by banks to determine risk, the VaR models weren’t taking into account the expected value of potential losses on that small one percent chance that something would go wrong, namely if housing prices started to decline. Management had so much faith in this type of forecasting and risk-management that they developed cognitive biases of relative optimism and underestimation of uncertainty. 

What does all this have to do with supply chains? Namely that while global supply chains strive to become more and more resilient, the advent of climate change is making it so that current supply chain forecasting based on past performance may not necessarily be a good predictor of what’s to come.  As the level of atmospheric CO2 steadily increases to become unlike anything seen since humans began keeping records the level of uncertainty in predicting changes in climate and extreme weather increases as well.  This uncertainty will have a major effect on supply chains, largely due to the multiplier effect of extreme weather events.  While more and more companies are planning for climate change along their own supply chains, collaboration across industries, sectors, and different stakeholders will become more important as a means of dealing with broad impacts. For example, PepsiCo has expanded supply chain management to include the use of climate risk tools in addition to its efforts to conserve water at the field/local community level, both vital steps to address risk and ensure the sustainability the supply chain.