Unifying The Supply Chain With Machine Learning Organized Information, Increases Speed, And Improves Efficiency Posted on : Jun 30 - 2020

Supply chain management has always been one of the most complicated business processes. There’s the complexity of multiple systems internal to a corporation, including accounting, manufacturing, inventory and more. Then there’s the need to share information up and down the chain. It was a problem before computers, in the mainframe days, and still exists as a challenge today. Machine learning (ML) has begun to have an impact on the supply chain, and it’s overdue.

Let me begin by pointing out that I am talking about an ML definition that isn’t limited to artificial intelligence. There’s certainly AI in areas such as natural language and some predictive arenas, but the inclusion of complex statistical analysis provided by procedural algorithms also provide insights that mean inclusion in ML.

In order to understand the broad opportunity for ML in the supply chain, we need to look separately at the two areas mentioned above.

Supply Chain Within The Organization

Back in the 1980s, I worked on a manufacturing company’s inventory system. The folks who built it only talked with accounting. It was great for accountants, but the user interface and data available were almost useless to the inventory people – and that was in a single system. The problem has only become exponentially complex.

Data silos. IT professionals have been complaining about data silos for decades, and they’re still a problem. The goal of a consistent, complete, corporate view of data is still just that, a goal. ERP, CRM, and other systems still have multiple, redundant data items with different data types. Manufacturing data doesn’t fit accounting data which doesn’t match sales order systems.

“In some ways, data moves more slowly than physical products,” says Rob Bailey, CEO, BackboneAI. “Breaking down the walls between data sources, and then aligning data to present a clear and accurate understanding of the supply chain is something that machine learning is addressing.”

Let’s use one example that is a necessary bane of product existence, the stock keeping unit (SKU). The SKU is a code to identify a specific type of product, and every company creates to track inventory. Note that “every company”, which we’ll discuss in the next section. For the purposes of an organization, the growth of departments, divisions, and national branches, even a single product can have different SKUs in the multiple systems within an organization.

Mediating between multiple systems can include tens of thousands of SKUs, so identifying similar products is something that can go much faster with ML. Natural language processing (NLP) is useful for the rapid scanning of product descriptions and then probabilistic categorization can link separate SKUs to provide an overall picture of a single product. This can speed the creation of corporate metadata and the ability to provide a global picture of products. View More