TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

Avoid Supply Chain Tsunamis Using AI to Read the Ripples

Managing a modern supply chain requires new architecture to enable AI and machine learning processes.

• By David Sweetman
• January 16, 2020

There's an old saying: "Red sky at night, sailor's delight. Red sky at morning, sailor take warning." If only such a barometer existed that would warn of any incoming supply chain disruption.

For Further Reading: Leveraging Data to Support Your Supply Chain Empowering Everyone to Make Decisions with Confidence IoT and the ML Connection

Perhaps one already does.

There is plenty that can interrupt the calm reliability of global supply these days. Brexit, trade wars, oil prices, wildfires, and economic fluctuations all add an element of unpredictability to supply, prices, or resources.

The potential for supply interruption is always present. Companies have used many strategies to protect against fluctuations (without buildup of excess inventories), including multiple sources of supply, contract commitments, currency hedging, and highly connected technology that can alter supply with fluctuations in demand.

Predicting Risk Through Artificial intelligence

Sensing supply chain risk is becoming easier with the help of insights and actions resulting from artificial intelligence and machine learning looking for changes that were sometimes too small to predict before. Systems need to become even more predictive across every element of the supply chain -- from the raw materials through distribution, production, delivery, and service.

Supply chain resilience is something that is now being built into products all the way from the design and selection of materials through to planning, manufacturing, delivery, and operation. It is only by tightly synchronizing the design-to-operation cycle that resilience can be built.

Financial markets can give us some view into prices and trends, enabling us to hedge our bets on longer-term supply positions, but we need intelligence even closer to the production. Delivery and operation need intelligence as well, such as using machine learning to predict stock issues before they become problems.

Managing for the Ripple Effect with Predictive MRP

Material requirements planning (MRP) systems have gone a long way to look at forecast demand, actual demand, current and expected inventories, and commitments so that all the dots are connected and promises to customers and production processes can be made reliably.

Although having the right component products to build a finished product is important, having the available capacity, tools, machines, and people to convert the components is equally important.

The essence of predictive MRP is the ability to provide insight into production capacity and material requirements together at the same time to meet demand. This removes the limitations of separated infinite and finite planning systems of the past that had to be run separately many times. Predictive MRP can visualize and solve finite capacity constraints.

It's also not enough to be planning on the top level only -- a chain is only as strong as its weakest link. All materials need to be available and synchronized at the right time. This means that planning systems need to see deeply into the multiple levels of assemblies, components, and subcomponents all the way through the bill of materials.

Speed, Intelligence, and Connection Make It Happen

Think of all the stuff we used to do without high-speed connectivity, such as balancing the supply chain and making adjustments in the last mile. We needed big inventory reserves to manage the fluctuations, requiring big warehouses with more transport and logistics, resulting in higher cost.

As connectivity enables speed, our ability to use technology to overcome the predictability gap becomes imperative. Our systems have to be more intelligent, work in real time, and be responsive. Living in the moment and not waiting for weekly runs, we need to predict and adjust faster in order to make decisions in real time that appear "natural."

Deeply connected, predictive, and high-speed supply chain systems now optimize the operations to make it happen. They require a new architecture that is lightning fast -- using memory rather than disk processing and using artificial intelligence to optimize decisions that are too fast to make on our own. These AI systems keep a finger on the pulse of the business and make predictions with machine learning.

The fundamental architecture is different. It has to work together and be complete; it is no good having inflexible long-term contracts from vendors and high variable demand from customers. Everything needs to work in unison together.

We can't wait to see how these new architectures will change the world.