What is a digital twin, in plain words?
A supply chain digital twin is a model of your real supply chain that is kept up to date with your own data and used to decide what to do next. It represents what matters to the decision, such as suppliers, sites, stock, routes, lead times and costs, and it lets you try a choice on the model before you try it on the business.
The term is used loosely, so a careful definition helps. A 2024 report from the US National Academies of Sciences, Engineering, and Medicine defines a digital twin as "a set of virtual information constructs that mimics the structure, context, and behavior of a natural, engineered, or social system (or system-of-systems), is dynamically updated with data from its physical twin, has a predictive capability, and informs decisions that realize value" [1]. It adds: "The bidirectional interaction between the virtual and the physical is central to the digital twin."
That gives four tests. A digital twin:
- Mirrors the structure and behaviour of the real system.
- Updates with data from the real system.
- Predicts what is likely to happen next.
- Informs decisions, and the decisions feed back into the real system.
Researchers draw the line in a similar place. In a review of digital twins in manufacturing, Kritzinger and colleagues separate a digital model (no automatic data exchange with the real object), a digital shadow (data flows automatically from the real object to the model, but not back) and a digital twin (data flows automatically both ways) [2]. They found that publications on the full digital twin were "scarce", while far more described models and shadows.
How is a digital twin different from a simulation or a dashboard?
The same National Academies report states it directly: "A digital twin is more than just simulation and modeling" [1].
| What it does | What it lacks | |
|---|---|---|
| Dashboard | Shows what is happening now, from live or recent data. | Does not predict or test options. |
| Simulation | Predicts what could happen under assumptions you set, usually for one study. | Not kept in step with the real operation. |
| Digital twin | Stays updated with real data, predicts, and is used for repeated decisions. | Costs more to build and maintain, so it needs a decision that justifies it. |
Many products sold as digital twins are dashboards with a 3D picture, or one-off simulations with a new name. Neither is wrong, but neither is a twin. In the terms above, a live dashboard is closer to a digital shadow [2].
Where twins are real, the uses are concrete. A review of industrial digital twins by Tao and colleagues covers applications in product design, production, and prognostics and health management [3]. For supply chains, Ivanov and Dolgui describe a digital supply chain twin as "a computerized model that represents network states for any given moment in real time", and argue that the disruptions of the COVID-19 pandemic showed the need for twins that map supply networks and give end-to-end visibility [4].
Does a digital twin need real-time data?
Not necessarily. The National Academies report stresses that the model should be "fit for purpose": its model types, fidelity and resolution should be chosen "to fit the particular decision task" [1]. In practice the update rate should match the decision. A port that reacts to a storm needs data refreshed in hours. A quarterly sourcing review can run on monthly data. Building minute-by-minute feeds for a quarterly decision adds cost without adding value.
When is a digital twin worth building?
A digital twin pays off when three things are true:
- The decision repeats. Inventory levels, supplier allocation, routing and maintenance schedules come back every week or month.
- Mistakes are costly. A stock-out, a missed shipment or a stranded asset costs far more than the model.
- The inputs are uncertain. Demand, lead times, prices, weather or supplier reliability vary, so the answer changes with conditions.
If any of these is missing, a simpler tool usually does the job. That is the honest test, and it is the first conversation we have with clients.
Can you trust what a digital twin tells you?
Only if it has been checked. The National Academies recommends that "verification, validation, and uncertainty quantification (VVUQ)" be "an integral part of new digital twin programs" [1]. In plain words: confirm the model is built correctly, compare it with what actually happened, and show how uncertain its predictions are. A twin that gives one confident number for an uncertain future is hiding something.
This is where a twin connects to decision support under uncertainty. Instead of asking the twin for one forecast, you run each option across many possible futures and look for the choices that hold up, an approach known as robust decision making [5]. We explain it in Robust beats optimal.
How do we build digital twins at Building Resilience?
We start with the decision, not the technology. We agree which decision the twin must support and who will use it, model only what that decision needs, connect the data that already exists, and test the model against history before anyone relies on it. Where key information sits in documents, such as invoices and packing lists, MaiMai, developed through our AI lab ClarityAILab, can turn it into structured data. The steps are set out on our digital twins page.
The methods behind it come from our own work: stochastic optimization enhanced by genetic algorithms for multi-step, multi-country supply chains, and an early warning algorithm that identified a significant spike in deaths in early January 2020, weeks before the first official COVID-19 case was reported in the UK.
What to do next
Write down the one decision you would most like to test before making it, and the data you already have about it. If that decision repeats, is costly to get wrong and depends on uncertain inputs, it is a candidate for a digital twin. Talk to us and we will tell you plainly whether a twin, a simpler model or a dashboard fits.
References
- National Academies of Sciences, Engineering, and Medicine (2024). Foundational Research Gaps and Future Directions for Digital Twins. Washington, DC: The National Academies Press. doi:10.17226/26894
- Kritzinger, W., Karner, M., Traar, G., Henjes, J., & Sihn, W. (2018). Digital twin in manufacturing: A categorical literature review and classification. IFAC-PapersOnLine, 51(11), 1016–1022. doi:10.1016/j.ifacol.2018.08.474
- Tao, F., Zhang, H., Liu, A., & Nee, A. Y. C. (2019). Digital twin in industry: State-of-the-art. IEEE Transactions on Industrial Informatics, 15(4), 2405–2415. doi:10.1109/TII.2018.2873186
- Ivanov, D., & Dolgui, A. (2021). A digital supply chain twin for managing the disruption risks and resilience in the era of Industry 4.0. Production Planning & Control, 32(9), 775–788. doi:10.1080/09537287.2020.1768450
- Lempert, R. J., Groves, D. G., Popper, S. W., & Bankes, S. C. (2006). A general, analytic method for generating robust strategies and narrative scenarios. Management Science, 52(4), 514–528. doi:10.1287/mnsc.1050.0472
Questions people ask
What is a supply chain digital twin?
A supply chain digital twin is a model of your real supply chain that is regularly updated with your own data, can predict what is likely to happen, and is used to make decisions. It represents suppliers, sites, stock, routes, lead times and costs, so you can test a decision on the model before acting on the real operation.
What is the difference between a digital twin and a simulation?
A simulation is usually built for one study and run on fixed or assumed data. A digital twin is kept up to date with data from the real operation and is used again and again for decisions, so information flows both ways: from the operation into the model, and from the model back into what the operation does.
Do you need real-time sensor data to have a digital twin?
Not always. The model only has to be updated as often as the decision needs. A weekly sourcing decision can run on weekly order, stock and supplier data; minute-by-minute data matters only for decisions made minute by minute.
How do you know if a digital twin is worth building?
Start from a decision that is made often, is costly to get wrong, and depends on uncertain inputs such as demand, lead times or disruptions. If no such decision exists, a report or dashboard is usually enough.