Why Digital Twin Projects Get Stuck at the PoC Stage
From a Single Project to a Standard
Key Facts at a Glance
Nearly half of all industrial companies in Germany use digital twins, but many projects never make it past the initial pilot phase. Why is that? How did Unilever, BMW, and Mars make the leap to full-scale operation, and what factors make scaling up affordable?
The Pilot Program Is Working. Now What?
The first digital twin is almost always a success. One production line, a dedicated team, a few custom-built interfaces—and after just a few months, a dashboard is predicting malfunctions. Then the second production line is supposed to follow, and it costs almost as much as the first. By the time the third plant is built, the budget is exhausted.
This is not an isolated case. 48 percent of German industrial companies with 100 or more employees use digital twins in production. Nevertheless 60 percent see themselves as laggards, while only 3 percent consider themselves leaders. A Capgemini study of more than 1,000 companies shows that technology alone is not enough.
Successful companies excel in six areas simultaneously: clear leadership and vision, digital work practices, collaboration with partners, intelligent systems, shared platforms, and integrated IT. Only 13 percent achieve this. This group achieves at least 65 percent greater results.
Four Reasons Why Digital Twin Projects Get Stuck in the Pilot Phase
No measurable goal: Unilever openly admits that in the early days not having defined clearly enough what measurable benefits the twins are supposed to deliver. Without a key metric, there is no solid business case for the rollout.
Lack of Data and Infrastructure: 57 percent of German industry lacks the necessary data, while 48 percent cite deficiencies in IT infrastructure as a common reason for failure.
Custom design instead of a blueprint: In the pilot phase, data models and interfaces are often created specifically for a single system. Every additional machine then becomes a new integration project, and costs rise proportionally.
No operational model: Who will maintain the model when the facility is retrofitted? Who will respond if data is lost? A digital twin with outdated data does more harm than none at all.
What Scaling Looks Like in Practice
BMW integrates building, facility, logistics, and vehicle data into a Virtual Factory for over 30 locations. The collision testing for new models takes three days virtually, instead of nearly four weeks of real-world testing. BMW expects production planning costs to be up to 30 percent lower.
Mars started 2020 with a single problem: overfilled packages at a plant in Illinois. Following successful testing, the digital twin solution was rolled out in the U.S., extended to the pet food business in Europe and China, and subsequently migrated to a shared cloud platform.
The pattern is the same in every case: first, solve a specific, measurable problem; then, replicate the process using a shared data set and platform.
Five Leverage Points for Driving the Economic Growth of the Digital Twin
Key metric before technology: scrap rate, OEE, or planning effort, with a baseline value established before the project begins and a designated person from the department in charge.
Standardized data models: The Asset Administration Shell (IEC 63278) describes assets using standardized submodels. In Use cases from Siemens, Festo, Bosch Rexroth, and other manufacturers showed reductions in engineering time and costs of up to 67 percent, and in product change notifications of up to 79 percent. The key factor is the underlying effect: Each new system uses the same submodels instead of creating a new interface.
A platform instead of a project: Data integration, model management, permissions, and security are set up once. Every additional digital twin is then a matter of configuration rather than a new development.
Plan operations from the very beginning: Responsibilities, updates, monitoring, and service levels should be included in the pilot phase, not just in Phase Three.
Get employees on board: Train teams to use the digital twin, and have recommendations initially reviewed and approved by people. Automated interventions should only follow once confidence in the results has grown.
Where Arvato Systems Comes In
Scaling doesn't start with technology, but with a measurable goal. Arvato Systems helps companies define appropriate metrics and provide transparency regarding the benefits, costs, and responsibilities associated with the digital twin.
Building on this, our solution creates a reusable database. It integrates data from production, SAP, MES, PLM, and service, harmonizes different structures, and makes them available through standardized data models and interfaces. This ensures that the next machine doesn’t become the next integration project. The pilot project serves as a blueprint for additional plants, locations, and use cases.
The approach also includes a robust operating model with clear responsibilities, quality standards, and monitoring. Finally, data rooms allow the digital twin to be extended beyond the company’s boundaries: manufacturers, suppliers, and service partners can provide the necessary information in a controlled manner without relinquishing control over their data. In this way, a single solution becomes a scalable foundation for collaboration across the entire value chain.
Do you have a pilot program that’s working but not scaling? Get in touch with us. Together, we’ll assess which components can be reused and what the realistic cost of the next location would be.
Written by
As Head of Strategic Business Development, Johannes Fuhrmann manages the portfolio and product development for the manufacturing industry, focusing on digital twins and data ecosystems. He previously worked at Deloitte and VELUX. He holds an MSc in Information Systems Management (University of Warwick) and a B.A. from HAW Hamburg.