The concept of supply chain digital twin has become very popular in recent years. It is essentially a digital model of your supply chain intended to be the foundation for companies to plan and operate their supply chain digitally and autonomously. The ingredients for a digital twin are: Data, Modeling capability, Algorithmic techniques that can take into account optimization as well as uncertainty. There is also one more attribute of digital twin. That is, like any other twins, the ability to change and grow together. As the physical supply chain changes then the digital supply chain needs to change with it.
This latter property implies a paradigm shift in supply chain planning. It means a shift from deterministic planning to probabilistic planning, from static modeling to dynamic representation, from pre-defined algorithms to adaptable algorithms and from periodic planning to event-driven planning. If your supply chain model is not changing then you have a digital picture of your supply chain not a digital twin. A digital twin is alive a digital picture is dead!
Let us look at these ingredients in more detail. Data comes from many different sources from transaction systems, IOT, streaming data, social media and so on. Having only planning data leads to static modeling. A simple example is supplier leadtime. This may change over time or it may vary by season. Having a static value for the model leads to inaccurate plans over time. Data is not used just to build a supply chain model; it can also be deployed to detect trends and patterns that are evolving. Furthermore, data provides the basic elements needed to find causes of events and for prediction of future events. To this end, data from every source can be relevant and needs to be explored to gain the insight needed. The ML algorithms have the capability to observe such underlying trends, and then invoke prescriptive solutions, or simply ask for help from users, to resolve issues. The third ingredient mentioned above is the capability of the model to represent the supply chain accurately on an on-going basis. An example to the contrary is spreadsheet planning, or a very rough and static model of the supply chain as represented by S&OP models.
A digital twin of the supply chain is a mirror image of every object in the supply chain: a supplier behavior, an equipment capacity, set-up times, operators, the dependency of customer orders to each supplier and millions more. Attributes and attribute-based planning is therefore essential to represent every property of supply chain objects. Products have attributes, WIP inventory has many attributes as it comes out of a process, quality and precision of an equipment are attributes, even carbon footprint of a supplier or a site is an attribute. Thus, attributes are used to create digital twin of the supply chain and be able to monitor their changing patterns. For example, the cost or efficiency of an equipment may vary depending on the season or usage.
In summary, ability to create an accurate and adaptable model of the supply chain and be able to deploy algorithms for prescriptive purposes is predicated on systems that can perform Sales & Operation Execution (S&OE) as well as S&OP. Model accuracy depends on attributes and availability of comprehensive data as the foundation that is needed for intelligent algorithms to create a digital and an autonomous supply chain operation that is constantly adapting.