Supply Chain Planning – If Ain’t Real-time Why Bother?
Most supply chain leaders would agree that supply chain disruptions are not exceptions anymore and they have to be considered as part of the plan. In fact, the so-called antifragile planning implies that exceptions and disruptions should be considered in the plan such that we benefit from them; and not just be resilient to them. As exceptions become more frequent on the demand side as well as supply side, the frequency of planning should be adjusted such that either we can predict events, desirable or not, and be able to respond to them in real-time. Most current technology providers have major flaws in their approach. Their assumption is that we first plan based on a deterministic model every week or every month, then if an event occurs, we deal with it manually or otherwise. Well, the original plan is flawed to begin with not just because it is deterministic, but also the plan has no idea what the conditions are on the ground in real-time. To this end, commitments are given, and allocations are made not knowing the extent of the current issues in the operations.
Thus, there is a separation between plans and operations. The conventional wisdom is “build a bridge.” As we all know, this approach will create both decision and data latency. By the time planning is aware of operations, or operations know what the plans are it is too late. A ship stuck at the port, a late truck, a tornado, or the closing of the Suez Canal are not all predictable. But by the time you find out and try to do something about it manually, it would be too late. It is also humanly impossible to know what your choices are if you must reallocate and change execution to make sure all your business priorities are optimally satisfied. A task almost impossible to accomplish optimally by manual intervention.
The ingredients of antifragile supply chains are:
- No separation between planning and execution. i.e. shared data in real-time
- Ability to respond and adjust the plan in real-time. Not manual planning or scenario analysis.
- A complete model or representation of the supply chain, i.e. a true digital twin. Representing resources in buckets and aggregation is not good enough. It is no different than spreadsheet planning.
- Intelligence to optimize, predict and respond, i.e. learn from the environment. The learning should be done by self-correcting the model on an on-going basis, self-improving the business policies and self-optimizing the algorithms.
By having continuous learning, one can monitor certain events and “vaccinate” the supply chain not to be vulnerable to such occurrences. One example would be the prediction of tornados in Southeast of the US in certain months of the year. Several decisions can be made to benefit from such predictions depending on location of suppliers, customers, distribution centers etc. in that region. Building inventory (i.e. predict and plan) can make you resilient for this type of event, but being able to respond fast enough when and if the event occurs or does not occur as predicted can make you more competitive and antifragile.
Even today with most supply chain planning systems, planners spend days if not weeks preparing plans. By the time the plan is made, and all the scenarios have been analyzed, the plan is no longer applicable since demand may have changed, conditions in execution are no longer the same or some other exception or disruption has occurred.
Having planning and execution in a unified environment enables planners to have plans which are optimally built almost in real-time. Thus, producing plans good enough for execution and potential disruptions. Planners, if so desired, within minutes they can improve the plan as needed or ask the system questions as to why certain decisions were made.
It is time to change the planning concepts of the 80’s when spreadsheets were introduced. We are now able to deploy much faster processes based on the technology which is available today and keeps rapidly changing. The approach of putting machine learning and/or GenAI and other recent innovations on old technologies and concepts of the past, may look good. But it is simply a whitewash of a technology that is decades old. The foundation is what matters, not just the façade.
