What-if scenario analysis is widely used by supply chain professionals in order to find the best response to an event in the past or one that is likely in the future. Given the number of variables and the millions of options that are available, only a handful of scenarios is examined because it takes too much time and effort to do more. It is a process by which a solution needs to be constructed out of many possibilities. Even when a satisfactory solution is found, it is unclear if all the consequences of the selected scenario is known.
For every decision made there are many consequences that may or may not be visible. For example, use of a substitute product has a cost associated with it and its use might deprive a high priority order from being on time at a later date and potentially loss of a customer. Understanding all these options and consequences of each decision in totality is simply overwhelming for any human being. However, it is a practice that has been popular for a long time.
We are now in a different era for a number of reasons including faster processing power, availability of large amount of memory and the use of ML algorithms that were perhaps available before but not practical because of lack of computing and memory power. Current technology makes it possible for the system to run millions of scenarios almost in real-time and offer a selection of solutions that the users can choose from depending on their objective(s), rather than spending so much time constructing one. There is always more than one objective to be satisfied, hence one solution may not necessarily be the “best.”
In order to accomplish scenario analysis automation, the system needs to understand the choices and possibilities. For example, use of substitute parts or alternate suppliers, or soft constraints such as some orders maybe slightly delayed. Using a true supply chain digital twin that understands the physical supply chain and is kept up to date, system uses clever algorithms to eliminate all the undesired solutions and find the best ones to offer to the users to choose from. One such technique is Constraint Propagation (CP) used widely in AI in order to shrink the search space and make the problem manageable enough to find a solution quickly. An example of this approach can be found in 8-Queen problem solution. To learn more on this and other topics in AI and ML used in supply chain planning click Here.