Why forecasting the supply chain is not like predicting the weather

Machine learning in effect today means more money and more resources for your business tomorrow.

My colleagues and I like to joke that if supply-chain professionals got away with what weather forecasters get away with, supply chain would be the most fun profession in the world. Forecasting automotive and advanced manufacturing supply chains today involves working with machines in ways that have never been done before.

Not utilizing machine learning as part of your supply chain today is like using 20th-century tools to do business in the 21st century. Machine learning is one of the most vital Industry 4.0 capabilities for decision-making, business planning and execution by executives.

Industry 4.0 is an integration of the physical, biological and digital world, built on a base of emerging technological breakthroughs, including autonomous capabilities, artificial intelligence, the Internet of Things (IoT), next-generation wireless technologies, nanotechnology, big data, blockchain and cloud computing. Some of this technology is in use today — on a mass scale, or just emerging and being tested. Others are in their infancy and not widely adopted or understood, such as blockchain and cutting-edge autonomous capabilities. Machine learning is being used now and can help your business today.

About machine learning

Statistical forecasting using software applications and computers has been around for decades. What’s new is that software can now learn from these statistics and build new models and scenarios with little to no human intervention. This is machine learning. To summarize machine learning in one sentence: it is the ability of software programs to perform specific tasks without using instructions and instead relying on models and inference to solve problems. The software consistently retrains itself based on the most recent sales data available.

There is so much volatility and change in global supply chains today that without machine learning, it’ll be difficult to forecast demand accurately. With less accurate forecasting, more time, money and resources will be spent on course correction, rather than innovation and new product development. Machine learning in effect today means more money and more resources for your business tomorrow. Original equipment manufacturers (OEMs) that can utilize machine learning will have better plans, resulting in fewer inefficiencies and less “firefighting.”

Why machine learning matters

Machine learning can incorporate all the internal data available, as well as external factors and leading indicators, to create a much more sophisticated forecasting model. This model is more accurate, more responsive to changes and has less bias than humans do. All of this results in a better-quality forecast, which enables more effective planning of supply chains — planning raw materials and supplies in addition to planning production capacity and human capital resources for production. Machine learning also enables better collaboration with suppliers so that the suppliers can better manage their own supply chains and be more responsive to you.

Most companies are still using traditional methods to review data, such as historical averages and run rates for demand forecasting. This process is cumbersome, with many steps, yet it still results in an inaccurate forecast that doesn’t take advantage of available data inputs. Machine learning significantly reduces the effort to create a good forecast and ignores the data that does not truly impact the product demand — it filters out the noise from the signal. Because machine learning-based forecasting is more responsive, it can react faster to early signals of changes and disruptions in the marketplace. It also typically has significantly lower bias in the forecast. Using IoT and the cloud, data can be integrated into forecasting models. Machine learning is not generic. It can be tailored to every company, every industry.