The field of artificial intelligence and machine learning comes with intimidating terminology and lots of new data — big data, structured and unstructured, from the web, social media, news, internet of things (IoT), events and weather. Supply-chain professionals need to understand all of this data, and become expert at deploying new services, tools and processes to manage it. Such an ability is becoming foundational to modern supply-chain management.
The need for users to focus on data inclusion, definitions and quality can’t be emphasized enough. A.I. uses existing data differently, and often users are seeking out external sources to create an expanded view of their environment. It’s vital, therefore, to have in place good data resource management (DRM) tools that can assist users and I.T. with acquiring this expanded view.
In addition, as supply chains draw on other sources of information such as the IoT and unstructured data from the web, it’s essential to categorize it for use in analytic systems. Most of the world’s data is unstructured, and modern systems need the ability to incorporate it into their operations.
Modern DRM requires the following:
Knowledge and incorporation of modern supply-chain data. Data can exist in multiple states —temporal, streaming, raw, cleansed, filtered — and is constantly changing. For sensing and machine learning, it’s important to know the differences between them, and be able to assess their reliability. Modern data is both analog and digital. The latter comes from electronic data interchange, IoT sensors, radio frequency identification, and wireless, to name just a few of the possible formats.
Adoption of DRM tools. Systems need an expanded and flexible data dictionary in order to include and categorize all inputs, including catalog and product data, which need constant updating. A.I. and machine learning toolsets can assist with this task, in addition to aiding in defining and managing the many new sources of data.
Acquisition of data. Companies need to research the many curated subscription data services and tools that are available, in order to subscribe to and automate the feeds into their I.T. systems.
Database toolsets. Storing and accessing data often requires three types of databases:
Data lakes for raw data,
Application databases that are specific to the system, and
Data warehouses, with extracted data for analytics and reporting.
Due to the huge quantity of raw data, third-party cloud storage might be required.
Refining and updating of roles and responsibilities. The question of who’s doing what is often overlooked, which leads to poor communication and duplicated or missing work. These responsibilities should be established from the start, to ensure the effectiveness of work and eliminate misunderstandings that might emerge later. Key roles within the organization include data scientists, software engineers, data managers and supply-chain scientists, who are the ultimate owners of the data. End-users know the why of data, and both they and data-management specialists know how to define it. Programmers don’t really like this kind of work, so roles should be clear to allow I.T. to write and deploy applications.
Outlook
Data management is often a neglected area within the overall organization. A.I. and machine learning, along with the whole world of external data, have become a catalyst for companies to invest more in this area. To take advantage of the possibilities of greater connectivity, and cope with information coming from new sources, excellence in data management is required. To assist with their data-modernization initiatives, end users will turn to network providers, data services and technology tools.
