Uncertainty, Models, and the Chaos Out There

Future is dominated by uncertainty. In terms of supply chain management – my main field of application – it comes in terms of variation in lead times or demand levels, or in the form of severe events such as hurricanes which can severely disrupt a supply chain. Optimization can help to cope with uncertainty, but I often get the feeling that we don’t completely “get it”.

Mathematical programming models have been used in supply chain planning for about 50 years. The vast majority of these models are deterministic, and that has now been described as a major weakness. Sure, that plan is optimal, but if demand, prices, or lead times change, then it’s not optimal anymore. A deterministic model may not result in a more resilient or robust plan.

Of course, many models have also been proposed to deal with uncertainty. Stochastic and robust optimization is now applied to many problems, and thanks to the significant improvement of mathematical programming solvers, we are actually able to solve several of these models. These models are also based on several assumptions, mostly some form of characterization about what is uncertain. By reading – and sometimes writing – those papers, I feel the touch of classical modeling, where everything is in order, well-structured and controlled.

I’ve had the opportunity to talk with supply chain planners and logistics managers in the industry, and I get a very different feeling. It’s the Wild West out there – everything is moving fast and furiously, and people do their best to cope with change as it comes by. They spend the little time they have left into analysis and trying to anticipate what is going to be the next big shift. And it’s not because they lack skills or knowledge in planning. While interesting and important – long-term risk planning doesn’t help meeting your monthly targets, but integrating this last-minute new customer into next week’s supply chain planning does.

In fact, the only industries I’ve been in touch with that seem to do extensive risk planning are the energy and military sectors. It’s a natural fit:  they have the resources and the time to do extensive analysis, and the risks are so huge that it warrants the effort. North America should survive the destruction of a Wal-Mart distribution center, but a nuclear plant meltdown or a major crash of the power grid is another matter. For these industries, the methodologies developed in academia to deal with risk and uncertainty seems to be a good fit.

As for the supply chain manager, I’m not sure research has yet found the right set of tools to help him/her actually deal with the chaos out there. Academic notions of risk, dynamic models and planning under uncertainty fall short to describe what they are dealing with. Sure, I’d like to bridge that particular gap between theory and practice, but I believe we need different tools to begin with.


  1. Michael Pouy says:

    Marc-André, we in the defense (or defence) arena do not generally use mathematical programming to optimize inventory investment. This approach has always assumed too much is deterministic. Instead, we use Lagrangian approaches or marginal analysis models to optimize investment across a range of items. We take these models a step further by using equipment readiness goals instead of fill rate or backorder goals. We still rely on assumptions about statistical distributions of lead time demand, even though these distribution do not fit very well. Recently we have developed ways around those assumptions, and we are seeing remarkable results.

  2. Michael, thanks for sharing your experience. In the post, I mostly meant that from my experience energy and military do put far more effort in using any kind of models in risk management than other industries or fields. The types of models may actually vary; Canadian Forces made some use of stochastic programming for several types of planning under uncertainty. I can’t say for the whole industry, of course.

Speak Your Mind