Consider a scenario in which state government leaders have three generational transit projects they could invest in but enough capital to support only one. Each project carries distinct benefits as well as unique trade-offs. As they evaluate options, leaders struggle to balance the risks and rewards, given the many competing priorities across the state. Moreover, with each option having a high, irreversible capital cost, making a suboptimal decision is expensive. To navigate this uncertainty, leaders build a digital replica of each project that models all key factors, such as capital cost, potential transit modalities, citizen accessibility, carbon emissions, and energy consumption. They then simulate numerous population growth scenarios years into the future and analyze each option’s current and future impact. The outputs of these analyses clarify the course of action that maximizes societal benefit and the overall ROI, and the leaders confidently announce their decision.
Digital twins1 are powerful tools that can dramatically improve the speed and quality of complex decisions like the transit investment above. Recent advancements in computing power and data availability mean that digital twins can now be used to enable government leaders to build accurate digital replicas of real-world assets and systems. These digital twins can enable accurate modeling, scenario planning, and analysis of both the direct and indirect effects of a project—and some digital twins are already implemented and doing just that. Ultimately, they enable better, faster decision-making and a higher ROI. As such, digital twins can enable governments to implement a “dig once” policy, in which potential issues and bottlenecks are identified before a project begins, thereby minimizing disruption to affected populations and decreasing both costs and the risk of delays.2
The number of active large-scale public infrastructure projects of this type is growing quickly around the world as a result of several landmark investment programs,3 including the $1.2 trillion Bipartisan Infrastructure Law in the United States,4 the more than €800.0 billion NextGenerationEU COVID-19 recovery plan in the European Union,5 and the Indian National Infrastructure Pipeline, which contains 1,110 projects under development for a total project cost of more than $2.3 trillion.6 The success of investment programs such as these is essential because they form a cornerstone of governments’ response to myriad ongoing global disruptions. These disruptions include rising geopolitical tensions, stagnating economic growth, the accelerating impact of climate change, growing populations, and rapid developments in AI.
However, translating a major influx of funds into action can be challenging. Access to the right expertise7 and strong governance in capital planning8 are critical, but they are not enough. Generating a good ROI also requires a sophisticated approach to planning and execution. Historically, however, fully modeling the real-world impacts of complex capital investments has been too complex for standard modeling approaches, which typically use defined mathematical relationships capturing cause and effect and therefore lack the ability to handle the multiple interdependencies within complex systems. It is in complex decisions with numerous interdependencies that digital twins are most effective, so their value to government leaders can be significant.
This article focuses on maximizing return on capital-intensive infrastructure investments, though there are many other use cases for digital twins, including redesigning complex processes, aiding annual budgeting or acquisition decisions, accelerating product development through deep learning surrogates,9 and increasing the resilience of supply chains.
Successfully rolling out a digital twin can take significant investment and time, but the ROI through accelerated decision-making and efficient resource allocation can more than justify the cost. In our experience, digital twins have the potential to improve capital efficiency, accessibility of services, and operational performance of public sector investments by 20 to 30 percent.10 When mapped to millions or billions of dollars in expenditure, this payback can be significant.
Government leaders face unique challenges in making high-impact investments
Given their size and complexity, planning and implementing large-scale infrastructure projects is always a challenge. However, for public sector projects, these difficulties can be compounded and new challenges can be introduced.
Government infrastructure investment projects typically have to grapple with most or all of the following issues:
- Government investment planning can be manual, siloed, and highly localized. Maximizing returns on large capital investments requires a comprehensive understanding of the full system and interdependencies into which the investment is being made, but constructing this 360-degree view can be challenging in the public sector due to both the number of stakeholders involved and a paucity of formal coordination mechanisms.
- Investments are often large, high-stakes, and long-term. Government investments in operations and infrastructure, such as the construction of new rail lines in urban transit systems, are often large, multiyear efforts. The size and nature of these projects make effective project planning and implementation particularly crucial, especially because returns will generally start to come in only on project completion (that is, when the public starts to purchase tickets for the new services).
- Public–private capital formation dynamics can be complex. Government investments will sometimes need to be combined with private and philanthropic capital to yield the desired outcomes, as is often the case in the construction of toll roads to mitigate urban congestion. The private sector may sometimes have more stringent requirements related to both the magnitude and time horizon of ROI, and it may place less weight on other outcomes, which may challenge government capital-planning approaches. In addition, decisions about where to invest across full, complex systems can require significant expertise and the weighing of multiple factors. This complexity can make it difficult to gain consensus on, for example, where to prioritize finite investment in transit infrastructure for a large-scale urban renewal project.
- Projects can entail multidimensional, dynamic problems with unanticipated indirect effects. The complex interdependencies between systems—such as between nodes of a defense industrial base or elements of the energy transition—can mean that well-intentioned actions have unexpected side effects, some of which may not manifest for some time. Increasing the speed of any single step within the operations of an asset may, for example, place strain on personnel working downstream of that activity.
- Decision-making can rely on instinct over data. Project complexity or siloed planning data can mean that the data required for fully informed decision-making is not available, not accessible, or potentially inaccurate. In our experience, there is a risk that leaders without access to sufficient data may place more weight on instinct and intuition in the decision-making process. The value at stake here could be significant, with the McKinsey Global Institute estimating that optimizing data and analytics across the public and social sectors could create approximately $1.2 trillion per year in value.11
- Investment processes can be manual, with long procurement timelines. Analog, time-intensive processes can create significant space for human error, potentially leading to delays or inefficiencies. For example, multiyear budgeting processes that flow across multiple levels of an organization or government can lead to rigid decision-making that makes it difficult to respond to changing circumstances, as well as to a lack of individual engagement or accountability.
Digital twins can be very effective in supporting decision-making
Given the scale and potential of government infrastructure projects—as well as the challenges involved in getting them right—data-driven, proactive, and accurate decision-making is vital.
Recent technological developments mean that digital-twin technologies can now provide exactly that: These products are already being used to drive visualizations, simulations, and optimizations of the physical asset to support decision-making.12 They allow the user to model “what is,” “what if,” and “what should be.” Digital twins typically consist of five elements that can be progressively built over time (Exhibit 1). Process or data flow maps and data models—the first two elements—are the foundation. Together, they make up a digital replica of a process, integrating the different streams of data and process information that need to be analyzed. These inputs may come from live data feeds, or they may be static, depending on the existing technology stack.
Image description:
A conceptual exhibit shows the five core elements—or layers—of a digital twin:
- Optimization layer: A secondary layer that, on top of simulation, maximizes or minimizes desired KPIs.
- Simulation layer: A layer that contains predictive capacity to estimate KPIs of interest when data models interact with process flow.
- Emulation layer: A layer that replicates the behaviors of physical systems or processes.
- Data models: A layer that contains abstractions of common objects and subprocesses that interact with the core process flow.
- Process or data flow maps: A layer that has the foundation sequence, layout, or structure underlying critical processes.
Of these layers, the simulation and emulation layers are the core focus of digital twin development.
End of image description.
The third element—the emulation layer—contextualizes the historical and real-time information (that is, seeing what exists in the system today) contained in the data model, enabling a bird’s-eye view or synthesis of the current state of a system (in other words, “what is”). This capability also enables quality control of the underlying model approach and assumptions—for example, through simulating known scenarios and comparing model outputs with historical data. This quality control is essential before the digital twin is used to look at future-state or “what if” scenarios.
The power of digital twins to significantly improve decision-making typically comes when models shift from emulation to simulation, which is the fourth element of digital-twin technology. With this element in place, the user can simulate what-if scenarios that assess the impact of different decisions or different futures on the overall outcome, thereby identifying “what should be.”
For example, consider a complex engine-maintenance process that requires a significant amount of equipment and personnel. Once the process is modeled in a digital twin, it becomes possible to quantify bottlenecks across the process instead of relying on intuition or one-off analyses. The digital twin also allows the user to see the results of possible future scenarios, such as comparing the impact of buying new equipment that could speed up a particular process step with the impact of hiring more personnel who could support multiple steps in the process.
Some digital twins also include a fifth element: the optimization layer. Rather than identifying the best interventions through trial and error, an optimization layer automatically recommends interventions based on previously identified criteria, which could include, for example, maximizing throughput or minimizing downtime. However, for some twins, the run time for a single model can be up to several hours, which means that availability of computational power can limit the ability to implement a first-best optimization layer. Running the model iteratively 1,000 times to identify the optimal outcome using a hill-climbing algorithm, for instance, could take weeks or months and may not be practical to inform decision-making. In these cases, the user may choose to run a finite set of scenarios through the model to find a course of action that meets predefined goals.
Overall, digital twins are intended to be reused over time to optimize investments and promote organizational effectiveness. They are therefore often considered a “system of innovation” within an organization and are a new paradigm shift alongside AI or machine learning models and gen AI tools (Exhibit 2). As such, digital twins are a key ingredient in future-proof technology architectures.
A conceptual exhibit shows the illustrative data infrastructure and integration platform that would constitute a future-proof technology architecture for an organization. The data infrastructure and integration platform consists of four stacked systems, each of which has 1 to 4 elements, which are represented by boxes within the layer:
- Systems of engagement, which allow business user interaction with analytics modules, including visualizations and workflows. The two elements are workflow and unified interface.
- Systems of innovation: An innovative end-to-end digital twin layer for simulation, optimization, and prediction. The three elements are AI and machine learning models (for example, forecasting); digital twins; and gen AI.
- Systems of differentiation: These are industry and company-specific operations management solutions for competitive advantage with basic simulation capabilities. The four illustrative elements included are integrated end-to-end planning solution, manufacturing execution solution, transportation management, and warehouse management.
- Systems of record: These are a transactional backbone for standardized core functionalities. The single element in this system is enterprise resource planning.
These four systems exist within a compute environment, which contains standardized infrastructure of data storage, access, and application integration.
End of image description.
How digital twins are already creating value for capital-intensive infrastructure investments
Digital twins provide the ability to emulate current states and simulate potential future states, offering a powerful tool for modeling the largest, most important, and most complex systems and processes. The most natural use case of digital twins may be to inform large investments in instances when traditional analytical modeling is not sufficient or when prior efforts to model or improve a complex process have not resulted in anticipated returns.
These new technologies can improve decision-making in a number of ways—including providing end-to-end visibility across a process or operation, creating a single source of truth, enabling rapid and low-cost scenario analysis, increasing decision-making speed, and mitigating the impact of potential side effects. As a result, digital twins are already being used by government agencies to guide capital investment decisions across multiple domains and at various scales.
Targeting capital investments to reduce cost and increase energy resilience for government facilities. A large government organization built a digital twin with advanced optimization and analytical capabilities, including capital project and power-sourcing optimization engines, to evaluate the highest-ROI investments in energy efficiency, cost savings, and resilience. After considering tens of thousands of possible investments, the organization identified a path to more than $100 million in potential savings while also improving resilience.
Providing end-to-end system visibility to improve urban planning in Australia. The state of Victoria, Australia, is integrating real-time data and advanced spatial technologies to create a digital twin of key elements of its physical and social infrastructure. The program aims to provide value by reducing red tape, improving government services, and building more-resilient communities through enhanced disaster response.13
Accelerating and optimizing decision-making for military force design. A government agency used digital-twin technologies to optimize investments in military capability in response to an evolving threat. The agency was able to accelerate force modernization by increasing the focus on priority projects and optimizing training schedules within the set budget envelope and desired mission outcomes.
Modeling global inventory to guide acquisitions and manage risks. Government leaders are using digital twins to guide at-scale procurement or acquisition efforts that are heavily influenced by nongovernmental entities. One government entity built a digital twin of aspects of its supply chain to model global inventory levels for key materiel, identify schedule and production bottlenecks, and proactively mitigate logistics disruptions in the system from, for example, the loss of a supplier.
Mapping complex operational interdependencies and identifying bottlenecks. Government leaders are using digital twins to inform investments in complex supply chains and operational processes in instances when analytical modeling is not sufficient to capture interdependencies and dynamics between stakeholders. One government entity built a digital twin to identify key limiting factors related to infrastructure and personnel within a highly complex process that had never been fully mapped.
Streamlining operations and reducing flight delays at a large airport. Leadership at an airport shifted from reactive to proactive planning by using a digital twin to model operations and to identify improvement levers and optimize investments to address rising flight delays. The model allowed the airport to balance the cost of additional staff and equipment with the operational improvement, leading to an overall 20 to 30 percent efficiency improvement.
While digital twins can significantly improve capital investment allocation in many ways and across multiple use cases, the common theme is an increase in ROI. Users of these digital twins can optimize throughput, cost, and safety months or years before buying new assets, investing in new capabilities, hiring more personnel, or breaking ground on new projects, saving precious capital and time.
Getting started
Building a digital twin requires a deep understanding of the process or system to be modeled, the technical expertise to aggregate data and represent a process appropriately in digital form, and the computational power to run and maintain the model.
While the details of the appropriate digital twin will vary across problem statements and organizations, the broad steps for developing and building the digital twin are generally the same (Exhibit 3).
Image description:
A conceptual flow exhibit shows the seven steps of a standard implementation approach for a digital twin.
- Establish mission-derived goals and where to focus: Define objectives and questions for the model. Determine if the tool is strategic (eg, investment decisions) or operational (eg, staffi¬ng optimization).
- Identify and prioritize use cases: Develop use cases addressing core strategic questions and sequence implementation based on impact, feasibility, and extensibility.
- Identify tech stack and data elements: Survey solutions (off-the-shelf, bespoke). Determine integration with existing systems. Map data elements for the data layer and use cases.
- Build data layer: Develop a cohesive data foundation. Map processes end-to-end where needed.
- Develop digital twin and simulation logic: Create logic to mirror real-world dynamics. Abstract ideas of extreme complexity if necessary.
- Test, refine, and analyze outputs: Test model against historical data. Analyze outputs to inform core decisions.
- Scale digital twin capability: Scale across the organization with the right data layer. Consider frequency of use to inform platform, visualizations, and development approach.
Feedback should be gathered from end-users throughout this development process.
End of image description.
Early steps focus on establishing the scope and goals of the model, as well as on building the data layer. While data availability, complexity, and cost can be an issue, a majority of data can often be pulled from enterprise resource planning systems and integrated with other external sources. Simplifying assumptions are often incorporated—stitching together imperfect data is typical—and a road map can be developed to refine the model over time as new data is collected.
Once the data layer is complete, the emulation and simulation logic can then be built out. Last, the model should be thoroughly tested for validity before being scaled across the organization as needed.
While successfully rolling out a digital twin can take significant investment and time, the return on this investment can be substantial, given the magnitude of current and forthcoming government investments in infrastructure and capabilities—and given the value in making effective decisions the first time around. Digital twins could offer governments a real opportunity to ensure that they are deploying their finite resources in a way that accelerates the energy transition, increases economic growth, builds resilience, and helps to navigate global disruptions.
