Digital twins, or virtual representations of physical objects or systems, aren’t a new concept in manufacturing. But many are underdelivering.
The issue isn’t the technology itself, though. It’s the scope. Some manufacturers are too ambitious, expecting their digital twin to solve every problem at once. Others reduce their digital twins to passive monitoring tools that generate insight but don’t drive action.
Starting with a specific business problem and layering data as you expand beyond the factory floor will create a more powerful digital twin that enables better, faster decision making.
The next wave of digital twins is about sequencing the right data, in the right order, to support the decisions that actually move the business. Here’s an overview of how to think about using various data sources to drive bottom-line outcomes.
Internal data: The foundation of scalable digital twins
To be useful, digital twins have to reflect the real world, which requires a lot of data. The first layer is internal data: ERP, MES, WMS, TMS, and sensor or machine data.
Without a solid internal data foundation, a digital twin isn’t a twin at all—it’s a model with blind spots. The intel from internal data informs staffing decisions, production planning, supplier coordination, and shipping strategy. It establishes a shared operational truth that teams can act on with confidence.
As manufacturers compile internal data, gaps often surface:
● Older generations of equipment may limit the amount or types of data available.
● Different manufacturers may follow different data standards, so data could be fragmented in plants with a mix of machinery.
● In-process machine updates (e.g., retrofitting older machines with new sensors) could delay getting the desired data.
Unaddressed, these issues can cause a digital twin to fail. The solution: be aware of these shortcomings and approach them as constraints to design around.
After shoring up their internal data foundation, manufacturers can amplify the impact of their digital twins with data from other sources.
Partner data expands insights beyond the plant
The next layer in a digital twin is partner data: supplier inventory, logistics data, delivery tracking feeds, etc.
This layer lets manufacturers be more proactive. For example, they can forecast disruptions before they hit operations, align production with material availability, and communicate changes across the supply chain ahead of time.
But acquiring partner data isn’t always easy. Trust and transparency concerns may prevent some partners from sharing detailed data with manufacturers. System incompatibility and uneven data maturity can limit what’s able to be shared.
The approach should be what is called “pragmatically ambitious,” where you help manufacturers figure out what kinds of partner data they can realistically obtain today, what might be available in the future, and where providers are needed as temporary workarounds.
Once a manufacturer understands its partner data, it can start considering how to apply real-world conditions to its digital twin.
External data brings the “real world” into the model
The final data layer in a digital twin moves beyond enterprise systems to environmental context. External data could include information on weather, infrastructure and traffic conditions, and public safety data.
External data is powerful because it reflects the conditions that manufacturers can’t control. The challenge is that external data is theoretically unlimited, but most use cases require only a narrow slice. For example, if your business operates domestically, you don’t need global weather data.
External data can serve many purposes, for example, where organized crime is a real business concern. For instance, one manufacturer used third-party data on road closures and police activity to inform its logistics routing, allowing for safer workforce planning and more resilient operations.
External conditions, such as law enforcement activity, affect every business. Feeding this data into a digital twin ensures its insights are as close to real-life as possible.
How to build digital twins iteratively without overengineering
Some manufacturers see the most success when they take an iterative approach to building their digital twin.
Start small, designing from the desired outcomes backwards and focusing on use cases that tie to specific business metrics.
The first iteration should be just complete enough to answer critical business question and just reliable enough to guide real decisions; anything more slows momentum without improving outcomes.
Starting small has two advantages. First, it limits the upfront investment needed to get to a functioning model. Second, it leaves room for improvement. As teams begin testing a digital twin, they often uncover new data sources that would further the twin’s accuracy.
Starting small doesn’t mean cutting corners, though. Digital twins need to be accurate in order to function as “twins,” but they don’t need to answer every question right off the bat.
Judge each iteration by whether it improves a specific decision or outcome, not by model completeness or data exhaustiveness. The signal to move forward comes from real user feedback and observed usefulness.
Early wins build manufacturing leaders’ confidence in the tool, and confidence is what motivates their use of digital twins to address more complex decisions over time.
AI amplifies the impact of digital twins
The most valuable digital twins don’t stop at the factory floor. They layer internal and external data to facilitate faster, more confident action.
These layers of data are crucial for manufacturers looking to implement additive technology like AI. When grounded in a solid data foundation, generative AI amplifies a digital twin’s impact, enabling faster decisions and automated action when parameters are clearly defined.
But these capabilities only deliver value when digital twins reflect how the real world actually operates. Then digital twins stop being passive models and start becoming active participants in how the business runs.
