The ultimate goal of a digital twin is to be a real-time, virtual replica of a physical product, process, or system. The digital twin approach is becoming a popular technique for addressing operational challenges. Paramita (Guha) Ghosh explains, “Digital twins can present a virtual replica of physical components, processes, or systems to enable improved understanding of the performance characteristics of such entities. With digital twins, industry operators can visualize, predict, and optimize the performance of individual components within a system or process, or the entire process from a remote location.” Within the consumer packaged goods (CPG) industry, digital twins have potential uses for optimization, simulation, decision support, root cause diagnosis, reporting, alerting, and serving as a data repository for training machine learning (ML) models with a much broader data scope.
Digital twins in the supply chain
Digital twins are implemented as a new infrastructure, frequently in the cloud, existing adjacent to the company’s operational infrastructure and attempting to be a near-real-time digital representation of the state of the operation. The digital twin stays in synchronization by continually obtaining “sensor” readings at various points along the operating topology and enriching these sensor readings with other data sources. Digital twin systems help create models of physical environments and commonly use spatial graphs to model and visualize relationships and interactions between people, locations, and sensors. They also come pre-packed with common machine learning techniques for analysis, and the ability to accept business logic code. In other words, digital twins are a convergence of the Internet of things (IoT), artificial intelligence (AI), and business analytics. Ghosh notes, “[The] IoT Has Catapulted the Use of Digital Twins. The IoT era has ushered in a seamless fusing of the ‘physical and the virtual worlds,’ offering a unique competitive differentiation to enterprises. According to expert estimates, 20 billion devices will be sensor-driven by 2020.”
Alicia Oriol, a Telecom Global Business Development consultant at ToolsGroup, notes, “[A digital twin] allows you to accurately test the resilience of a complex, multi-echelon, global supply chain.” Because supply systems are complex, she cautions digital twins are seldom exact copies. She explains, “They may reflect quite disparate aspects of the supply chain and products, ranging from supplier to service characteristics, and at many different levels of detail. But the extent and the thoroughness of the simulation depend on the importance of the information to the value chain. … Unexpected needs can lead to ongoing changes and additions.” At CPG companies, in addition to the virtual sensor readings of operational states provided by enterprise resource planning (ERP) and other systems along the supply chain, physical sensors can provide data from production line machinery, gateway sensors, vehicles, robots, computer hardware monitoring, climate control, and so on. In this way, the digital twin can integrate data across systems, external vendors, and physical devices to obtain a holistic perspective of the operation.
Leveraging digital twins for supply chain planning
Oriol notes one reason interest in supply chain digital twins is growing is their potential for supply chain planning. She explains, “You can think of a digital twin as the ultimate ‘what-if’ scenario planner.” Steve Banker (@steve_scm), Vice President of Supply Chain Services at Arc Advisory Group, isn’t so sure. When applied to supply chain planning, he writes, “The term digital twin sounded like hype to me.” Nevertheless, he notes, “The great majority of executives working for companies providing SCP solutions really like the term digital twin, although they did tend to define its key features somewhat differently. At the heart of a supply planning solution is a model of the supply chain process. The term digital twin may have a great deal of marketing puffery attached to it. But supply planning models are amazing. Using the term digital twin allows SCP executives to talk about SCP models, and how their companies’ approach to modeling is differentiated.”
Banker’s concern is that digital twins might not provide the timeliness and granularity some planning timeframes require. He explains, “Supply chain planning is done for different forecast horizons — factory scheduling and fulfillment planning may be focused on what will be made and delivered in the next week. This is known as operational planning. Tactical planning is focused on what will be made and delivered in the next month or next few months. Strategic planning is focused on even longer time horizons — plans going out many months or even years.” He goes on to note, “Toby Brzoznowski, the chief strategy officer at LLamasoft, … defines the digital twin somewhat differently than others in the industry. The digital twin is not the individual operational, tactical, or strategic model. It is a separate layer of the SCP solution stack that facilitates the strategic, operational, and tactical plans to operate from common data and parameters — everything set-up times, to lead times, to bills of materials, to shelf space available for a product at an individual store. In short, the digital twin is the ‘end-to-end reference model always available’ to operational, tactical, or strategic models. These different planning models than apply different statistical, machine learning, and AI algorithms, and different work flows, to answer different types of questions.” Banker concludes, “A good digital twin also gets data feeds that keep the supply chain model up to date. Further, the model is easy to configure and change.”
Oriol concludes, “Digital twins drive benefits such as improved business planning and improved sourcing, procurement and supplier management.” There are excellent use cases for supply digital twins, like supply chain optimization and new product introduction, but digital twin supply chain planning still suffers from growing pains. Some of the challenges faced when designing a digital twin strategy include:
- Intended use. A company must determine the intended uses of the digital twin so it provides enough value when complete, and focuses the scope of the data and sensor collection.
- Data collection. Finding the right balance of information depth to collect for the digital twin is essential. Collecting too much data or data at too low of a granularity/frequency can increase the cost of implementing the digital twin with little to no corresponding benefit.
- Locating appropriate sensor reading locations throughout the company. Reusing existing interfaces is the most cost effective to collect sensor data. However, it is likely some additional data will be required by the Twin to enable its full intended functionality.
- Integrating, normalizing, and interrelating the sensor data. Integrating, normalizing, and interrelating the sensor data at the digital twin is necessary so it can be used for analysis. Vendor systems provide the tools and enabling infrastructure for these activities, but it remains up to users to configure and develop the final model.
- Defining the digital twin’s business logic and functionality. Obviously, if the digital twin’s business logic and functionality fail to mirror real world operations, any insights it provides will be suspect if not useless.
- Creating the digital twin’s usability experience. This activity could involve applications, reporting, alerting, and the need for feedback interfaces into the operational systems. Vendor systems provide the tools and enabling infrastructure for this, but it is remains up to users to configure and develop the final model.
Creating the right model begins with well-defined goals to drive a digital twin project’s implementation.
 Paramita (Guha) Ghosh, “Fundamentals of Digital Twins,” Dataversity, 28 August 2019.
 Alicia Oriol, “Modeling Supply Chains with ‘Digital Twins’,” ToolsGroup Blog, 30 May 2018.
 Steve Banker, “Digital Twins In Supply Chain Planning,” Forbes, 2 January 2020