Supply chain digital twins might be one of the most exciting concepts for anyone affected by recent supply chain disruption. But as with many hyped technologies, closer scrutiny is warranted.
The use of digital twins to understand supply chains and predict effects from uncertainties such as COVID-19, trade wars and natural disasters does hold promise. However, digital twins are still in their infancy, and supply chain's complexity creates a number of obstacles to meaningful or widespread use of these digital replicas.
A digital twin for the supply chain
At its simplest, a digital twin is the virtual model of something physical, such as a thing, process, place or person. Well-known examples of digital twin uses outside of supply chain management include studying an aircraft engine's digital twin to understand safety risks and monitoring digital twins of industrial machinery such as oil rigs to predict when they will break down. A ground handling and in-flight catering service provider has even used digital twin technology to understand and improve food operations.
Perhaps less well-known is the use of digital twins in supply chain management.
"Although the term digital twin of the supply chain is being used more, it is still a confusing term for many," said Steve Banker, vice president of supply chain services at ARC Advisory Group, a management consulting firm.
Steve BankerVice president of supply chain services, ARC Advisory Group
Digital twin has a specific meaning, but in practice can lean toward an almost catchall term. Adding supply chain to that complicates the issue tremendously, because that term can refer to part of a chain or the supply chain as a whole, in other words as an end-to-end concept.
Some vendors capitalize on that confusion.
Technology suppliers can use the term digital twin to try to differentiate their modeling tools and hold the attention of their listeners longer than they would otherwise, Banker said.
One key is understanding what distinguishes a digital twin from a model or simulation, which a number of experts said is its ability to query data and truly learn something about the thing that it represents.
Experts weigh in on digital twins across industries:
A robust digital twin of the supply chain as a whole would mirror the many facets of that supply chain. It would need to use AI and data mining techniques and streaming data inputs from various sources, such as IoT sensors, and ERP, product lifecycle management and enterprise asset management systems to comprehensively model processes. Stakeholders could use this for tactical, operational and strategic planning across multiple tiers of the supply chain.
The long-term vision of a supply chain digital twin is a single integrated model that works across many processes within and even between companies for predictive and prescriptive modeling across the organizations. But bringing that vision into reality is challenging. Creating supply chain digital twins requires intensive work to improve data aggregation and data modeling and aligning these with the appropriate kind of analytics. Yet, even now, there are more targeted ways to use supply chain digital twins that are yielding benefits.
Challenges and speculation
A supply chain is, by its very nature, complex.
"The idea of a supply chain is not one process, but many processes," said Michele Pelino, principal analyst at Forrester Research. "The value of a digital twin is that you can model your whole supply chain virtually."
This would enable stakeholders to run various scenarios to understand issues and weaknesses in the supply chain.
For example, this would mean getting to a point where digital twins work across different kinds of processes, such as simultaneous updates across inventory management, warehouse management, fleet management, track and trace and cold chain monitoring.
3 layers of supply chain functionality
Three main layers compose a robust supply chain digital twin, according to Tim Payne, analyst at Gartner.
- The data layer consumes, organizes and models data from business applications like ERP and IoT platforms. It also provides master data management across this data. This goes beyond simply building a data lake or even a data warehouse because the data must be properly cleaned and organized for the digital twin.
- The data association layer sits on top of this and forms the basis for the supply chain digital twin. Each enterprise will want to standardize on one provider for this layer so that it operates as a comprehensive supply chain digital twin across the organization.
"I think we will see vendors specializing in pulling the data they need from a data lake and then do the work of associating the data in the appropriate way," he said.
- A predictive and prescriptive analytics layer sits on top of the data association layer. Many organizations will run standalone analytics software from different vendors on top of a single data association layer.
"The leading enterprises want a choice of predictive and prescriptive analytics to plug in," he said.
Down the road, companies will roll out these kinds of capabilities using a composable architecture build on top of a collection of microservices for these different capabilities.
"Every supply chain is unique, and even in the same industry, the patterns in their data is different," Payne said.
Supplier trust needed before integration
A comprehensive supply chain digital twin would need to integrate data from multiple tiers of a company's supply chain, such as the enterprises that sell the products directly, their suppliers and the suppliers' suppliers. This kind of modeling makes it possible to detect the potential chokepoints and other issues.
"There are many challenges when products and processes outside the boundaries of the organization are concerned," said Vishnu Andhare, consulting manager of engineering services at ISG, a technology research and advisory firm.
How one automaker relies on digital twins:
For example, building relationships that allow for a flow of information across company boundaries is no easy feat, Andhare said. Integrating information across various technology silos is also a challenge for both business and technology leaders.
A sophisticated supply chain digital twin will also need to integrate data from a variety of business applications, including manufacturing execution systems, ERP and supply chain-specific tools, Andhare said.
Another integration challenge lies in sharing data models across different types of use cases. Enterprise leaders will need to find ways to build a digital twin based on a consistent and shared way of representing a model. It will also need to support different kinds of business data, sensor data and contextual data, and participants in the supply chain will need to maintain sovereignty of their data.
That recognition is key: A multitude of stakeholders need to be involved to create a successful supply chain digital twin strategy.
Supply chain digital twin payoffs to come
Digital supply chain twins need to be considered part of a strategic supply chain framework to drive ongoing transformation from "chain" practices and technology to digital supply "network" practices and technology, said Matthew Lekstutis, global managing partner and supply chain consulting lead at Tata Consultancy Services. In this context a digital supply network twin could make it easier to simulate the different disruptive outcomes and risk mitigation scenarios and the consequences of response actions on the enterprise and the ecosystem.
In the long term, enterprise leaders may look for ways to connect digital twins to each other to share data, improve overall performance and improve decision-making across organizations, Lekstutis said. Existing industry specific B2B commerce exchanges will provide the ideal backbone for supporting interoperability across the digital twins used by different trading partners. In the future, this kind of integration will make it easier for companies to respond to massive changes, such as those demonstrated in the aftermath of the pandemic.
Today's supply chain digital twin opportunities
At the moment, a fully realized supply chain digital twin is not some off-the-shelf technology. Partly that's because it would be many digital twins of specific aspects of the chain that are linked together. But while no one vendor sells a digital twin of the supply chain, many vendors are starting to sell early versions of the concept -- platforms or applications technologists or leaders can use to handle some aspects of strategic, tactical or operational models.
Supply chain digital twins are still early in the hype cycle, and it will be several years before they fully pan out, Payne said. In the meantime, there are tools and more limited versions of supply chain digital twins that can improve predictive and prescriptive analytics today. These can help supply chain leaders and others develop the skills and processes to effectively use more fully realized digital twins once the technology matures.
To find examples of digital twins in the supply chain currently in use and learn from them, supply chain leaders can look for specific and real-life examples in areas such as procurement, logistics or manufacturing. These are often tied up with overall digital transformation efforts and terms such as digital supply chain and connected supply chain.
For example, Unilever uses digital twins of factories. Factory equipment and machines are connected with IoT sensors and intelligent edge services in the Azure IoT platform. They send data on everything from temperature to production cycle times to its digital twin software, which Unilever's engineering team created in partnership with Microsoft partner The Marsden Group. Both machines and processes are represented and stakeholders can mine them for patterns, using analytics and machine learning algorithms, which in turn help predict outcomes.
Using the digital twins has helped the company improve on products and increase production capacity, due to new insights.
7 building blocks for supply chain digital twins
Seven building blocks are required for a digital twin of the supply chain, Payne said. Here's what he recommends.
- Real-time transactions and events need to be captured automatically from IoT, enterprise applications and third-party data sources to make updates to the model every couple of minutes.
- Entities, attributes and parameters that make up the supply chain need to be automatically extracted from the data streams. These can include things like products, customers, lead times, bill of materials and source of origin. More granular correlations lead to more precise models.
- Configuration and correlations indicating the relationship between entities, attributes and parameters need to be mapped out using machine learning algorithms to cluster and identify relationships. This could include things like which products connect with a given customer and facility and how these are affected by variations like seasonality.
- Probability distribution needs to be calculated based on this data. For example, a model might define a standard lead time and a mathematical expression for determining what influences it.
- Dynamic pegging enables dynamically changing the relationships between entities through the model. A predicted customer order may be pegged to a specified amount of projected inventory pegged to a manufacturing request. When constraints or opportunities affect the plans, dynamic pegging can adjust these connections. This makes it easier to predict and respond to the impact of major events like a regional earthquake that may affect multiple nodes of a supply chain at the same time.
- Supply chain managers need tools to apply rules of engagement for certain service levels and inventory turns. The digital twin of the supply chain needs to make it easy to represent the kinds of decisions managers might make that can be simulated on the model.
- Supply chain visualizations provide the ability to visually see the supply chain from different angles using different kinds of analytics.