While we are living in a world that is characterized by uncertainty, many planning solutions still assume that the future is characterized by predictable inputs. The way I see it, the consequence is a plan that suggests a future state that we know will never materialize. And when things do not turn out as had been planned, the forecast error or the longer lead time from the supplier is blamed, though we knew upfront that it was only the best possible guess at that moment. How can probabilistic planning for your supply chain help to alleviate this? Let me explain.
Uncertainty or variability has always been an inherent part of supply chain management. For many decades, practitioners as well as researchers have been aware of the uncertainty they are facing and have tried to develop strategies to deal with it. The most obvious source of uncertainty that I’ve witnessed in supply chain is the demand uncertainty. And most supply chain professionals do realize that forecasts are by nature incorrect and that depending on how reality will unfold, the outcome in your supply chain will be different. And the same applies for lead times, machine output, etc., which means that the supply chain environment in which we’re operating is extremely uncertain.
But, that’s not how our traditional planning systems are looking at the supply chain. They take one single value for a lot of inputs that are used to optimize that supply chain: the expected demand, standard lead times, average production rates, yield, etc. These values are not fixed at all. However, based on these assumptions, they generate a plan for which we know in advance the outcome will not be as the plan has optimized it. The traditional way to deal with these uncertainties is very much focused on setting optimal safety stock levels, but that may not necessarily be the best/cheapest/most effective way to deal with that uncertainty. I believe a better option may be to increase capacity during a certain period, to consider outsourcing, to use an additional supplier, etc.
The traditional answer to this uncertainty issue is scenario management. Some “experts” in the company are asked to define a few alternative scenarios of how the future might unfold. Very often, these are some simplistic assumptions that are restricted to assuming that demand for product group A will be some percentage points higher than foreseen; and for component B, there could be a shortage due to an issue with a supplier. Planners then need to align on how plans would be best set up in case these events occur and see what the impact is.
When we get to the point that these scenarios do get evaluated, the next problem is to translate this limited set of scenarios into a meaningful set of decisions. It doesn’t make sense to make some kind of “average of scenarios” and then base your decisions on this completely random average future. The average of the optima is not the optimum.
→ Implicitly, a lot of companies realize that scenario management has a lot of shortcomings, and that is also the reason why, in practice, very few companies have embedded scenario management as part of their operational planning cycles.
The reality is that not a limited number of scenarios may unfold in the future, but that—given the large number of stochastic inputs that a supply chain is faced with—a nearly infinite number of scenarios may unfold. The few scenarios that typically can be generated in a semi-manual way are by no means a correct representation of this myriad of possible future realities.
Given the complex interactions that exist between the different factors that influence the supply chain, the only reasonable way to gain insight into this dynamic is through simulation. By doing many simulations, you already get a better resolution on the picture of how the future may unfold, at least sufficient to base your supply chain decisions upon. Hence, by letting a machine automatically generate a very wide scale of scenarios, you avoid the problem of semi-manual scenario generation.
The major challenge that follows is to use this information to base decisions upon. This is precisely where probabilistic planning for your supply chain comes into the picture. The insights that the simulations give need to be translated into optimal decisions.
→ New metrics—like probability-to-execute—can help management gain insight in the delivery performance that may be obtained through a set of decisions in that uncertain world we operate in.
Read on below to discover more about probabilistic planning with a digital supply chain twin:
> The evolution of digital twins: Probabilistic planning
