In the course of doing a global study on the supply chain planning (SCP) market I talked to executives at solution providers from across the industry. One topic that came up over and over was the idea that a supply chain planning model was a digital twin. A digital twin for supply planning is a digital replica of the physical supply chain process. A robust digital twin is needed to enable a self-healing supply chain.
Supply planning provides a solid return on investment based on requiring less inventory to meet service levels, requiring less labor and energy to produce goods, and being able to deliver goods to customers at a lower total landed cost.
In previous article, I covered the idea that a good digital twin for supply planning involves the ability to represent the supply chain process at the right level of resolution. But several executives said while this was a good starting place, a good digital twin in SCP required more. Omer Bakkalbasi, the chief innovation officer at Solvoyo, wrote me that a solid planning solution needs “the ability to update the representation with transactions as the world changes. After all, a living digital twin is useful for decision support, not a copy that is frozen in time.”
The quality of the solution that comes out of a planning application depends upon the quality of the model. Supply planning models are based on a model that, at a high level, spans getting raw materials into a plant (with lead times), understanding how long it will take to make those goods, and understanding how long it will take to deliver those goods to customers. If the way a product is made changes – a change to the bill of materials (BOM), for example – and that is not reflected in the model, the ability of the plan to generate savings degenerates. Similarly, if lead times change, and that is not reflected in the model, the plans generated are less than optimal. The issue of model degeneration has long been an issue in this market.
LLamasoft Supply Chain Application
Some of the updated parameter data is generated internally. An example would be BOM updates. The large enterprise resource planning companies argue they are better positioned to update the planning models because there is an elegance and seamlessness to the integration that best of breeds can’t match.
But some of the data that needs to be updated is external – lead times would be an example. JDA Software is partnered with FourKites, for example, to get the real-time lead time data so critical to these models.
Mark Stanton, a senior director at JDA Software, refers to this as “dynamic parameter calculation.” The good news is that many suppliers in the supply chain planning market are specifically tackling this problem with new product development and alliances.
In some cases, machine learning is being used to improve the data inputs. AspenTech, for example, is using predictive analytic inputs on when key machinery in a refinery will break down to allow alternative production schedules to be generated in a more autonomous manner.
Machine learning is also seen as a useful way to move from one number parametrization – the lead time from this supplier in Asia to the plant in Oregon is 26 days, for example – to being able to use variable parameters more reflective of what is happening in the real world. For example, in October the lead time for the supplier is 31 days, in January it is 20. Product development is going on in this area. Cyrus Hadavi, the CEO at Adexa, has told me that they are providing this kind of capability for a couple of clients.
Mr. Stanton has also pointed to the what is occurring around digital supply chain twins to open new opportunities. A digital twin should “provide real-time visibility as well as the ability to orchestrate the supply chain network” with advanced supply chain control towers. Johanna Småros, the cofounder at RELEX Solutions argues that the “control tower functionality should be embedded where decisions are made (typically the supply chain planning and optimization software), that all well-defined exceptions should be managed autonomously by the system, and that only exceptions benefiting from manual attention and more persistent performance issues should be presented to the planners for action.”
The supply chain field is never short on jargon. The term “self-healing supply chain” is beginning to be used. This term reflects the idea that parameters should automatically update. It also includes the idea that the best plan is useless if unexpected events occur. It is important not just to create an optimum plan, but to be able to replan using a robust control tower as needed. A self-healing supply chain is impossible without a robust supply chain digital twin.