# Digital Twin in Manufacturing: Business Case, Data Foundation and Implementation

Author: Marin Niehues, Niehues Consulting · Published: 2026-01-01 · Updated: 2026-11-16 · AI Adoption & Operations
Source: https://www.niehuesconsulting.com/blog/digital-twin-in-manufacturing/

A digital twin in manufacturing is a virtual representation of a machine or production line that is continuously synchronized with operating data and prepares decisions on the shop floor. It pays off when it addresses a specific decision with measurable value, such as unplanned downtime. This article provides an overview and does not constitute legal advice.

A representative survey by Bitkom, the German digital industry association, among manufacturers with at least 100 employees shows the pattern: the majority consider digital twins important, while six in ten see themselves as laggards. The cause rarely lies in the technology. What is usually missing is a target metric with a baseline, and the data foundation is thinner than assumed.

**A digital twin is a tool for recurring decisions in the plant. Its business case holds when the decision it improves has a baseline in euros or hours.**

## What is a digital twin in manufacturing?

A digital twin in manufacturing is a model of a real asset or process that exchanges data with its physical counterpart in both directions and thereby represents and predicts its behavior. This two-way connection distinguishes it from a conventional simulation, which runs on fixed assumptions.

The concept goes back to Michael Grieves, who presented it at the University of Michigan in 2002 in the context of product lifecycle management. The term "digital twin" was coined by John Vickers of NASA and appears in a technology roadmap of the agency in 2010. The US National Academies of Sciences, Engineering, and Medicine state in a foundational report that a digital twin is more than modeling and simulation: the bidirectional interaction between the virtual and the physical is essential.

A widely cited literature review in IFAC-PapersOnLine classifies systems by their level of data integration and shows that many solutions labeled as twins sit at lower levels.

**Three levels of data integration**
Digital model: the model is fed with data manually. Example: a simulation model for layout planning.

Digital shadow: data flows automatically from the asset into the model, while feedback to the asset is applied by people. Example: a condition model that indicates wear and recommends maintenance.

Digital twin: data flows automatically in both directions. The model changes parameters of the asset or the schedule directly or through an approved rule.

For decision-makers, this classification matters more than the label. Many business cases can already be realized with a digital shadow. Full feedback only pays off when decisions are taken so frequently and so quickly that a person in the loop becomes the bottleneck.

## Where does a digital twin pay off in manufacturing?

A digital twin pays off in manufacturing wherever costly decisions are taken repeatedly under uncertainty, above all in maintenance and commissioning as well as in process control and capacity planning. In the Bitkom survey, 79 percent of companies expect a digital twin above all to enable predictive maintenance with fewer unplanned outages.

- **Predictive maintenance:** the model compares the expected with the measured condition of an asset and reports deviations before a failure occurs. Maintenance follows condition rather than the calendar.
- **Virtual commissioning:** control software is tested against a model of the line before the line is built. Errors surface on screen rather than during ramp-up on the shop floor.
- **Process control and quality:** the model shows the relationship between process parameters and scrap and proposes settings.
- **Capacity and energy planning:** variants of order sequencing or shift patterns are tested before they put load on the line.

The plant's bottleneck determines which application comes first. A site with high asset availability but long changeover times gains more from planning than from maintenance. How AI applications in the plant are prioritized overall is covered in the article [AI in production and maintenance](/blog/ai-in-manufacturing-and-maintenance/).

**Digital twins in German industry**
48 percent | of manufacturers with at least 100 employees use digital twins in production (Bitkom 2025, up from 33 percent in 2022)
57 percent | see missing data as an obstacle to adoption (Bitkom 2026)
48 percent | cite shortcomings in IT infrastructure as a reason projects fail (Bitkom 2026)
60 percent | rate their own company as a laggard on digital twins (Bitkom 2026)

**Exhibit 1: Almost half of manufacturers use digital twins, up from a third in 2022**

| Label | Value |
| --- | --- |
| 2022 | 33% |
| 2025 | 48% |

Source: Bitkom, Digitale Zwillinge in der Industrie, representative survey of manufacturers with at least 100 employees, 2025
Graphic: [SVG](https://www.niehuesconsulting.com/blog/digital-twin-in-manufacturing/grafik-1.svg) · [CSV](https://www.niehuesconsulting.com/blog/digital-twin-in-manufacturing/grafik-1.csv) · [JSON](https://www.niehuesconsulting.com/blog/digital-twin-in-manufacturing/grafik-1.json) · Niehues Consulting, [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/deed.en). Data: Bitkom, Digitale Zwillinge in der Industrie, representative survey of manufacturers with at least 100 employees, 2025; the source's terms apply.

## How does a digital twin pay for itself?

A digital twin pays for itself when the annual value of the improved decision exceeds the cost of building and running the model within two to three years. The calculation starts with the baseline, meaning what the decision costs today.

A worked example for a bottleneck line:

1. **Establish the baseline:** the line is down unplanned for 300 hours a year. Each hour costs 4,000 euros in lost contribution margin and extra shifts. That makes 1.2 million euros a year.
2. **Limit the effect:** only part of the downtime announces itself in measurement data. The failure analysis shows that 40 percent of the hours stem from wear on assemblies that can be monitored with sensors. That is 120 hours.
3. **Set the target:** the twin is to replace half of these hours with planned interventions. That gives 60 hours and a benefit of 240,000 euros a year.
4. **Account for all costs:** sensors and integration with controls and MES, plus model building and training. Add ongoing operation with model maintenance and licenses.
5. **Test:** if total costs over three years exceed three times the annual benefit, the scope is reduced or a different line is chosen.

A common error is to apply the benefit to all downtime instead of the share that shows up in data. Another is to omit operating costs, although an unmaintained model gives wrong recommendations after the first modification of the line. How to bring benefits and costs together robustly is shown in the article [AI business case: calculating benefits and costs robustly](/blog/ai-business-case-benefits-and-costs/).

## What data does a digital twin need?

A digital twin needs master data on the asset, ongoing condition data with timestamps and event data from production control. If one of these sources is missing, the model remains at the level of a digital model.

- **Master data:** the structure of the asset and its maintenance history. These are often scattered across several departments.
- **Condition data:** measurements from sensors and controls, such as temperature or vibration, at a time resolution that suits the question.
- **Event data:** orders and fault reports from MES and ERP. Only these link a measured value to the order and its outcome.

Established standards exist for structuring these data. The ISO 23247 series describes a framework for digital twins in manufacturing in four parts, from principles through a reference architecture to information exchange. IEC 63278-1 defines the structure of the Asset Administration Shell, the standard for the digital description of an asset developed within Germany's Industrie 4.0 initiative. ISO/IEC 30173 harmonizes the terminology. OPC UA has become the established link to the control level.

The data pipeline is the largest single cost item of implementation, which matches the obstacles named in the Bitkom survey. How to assess data quality before starting is described in the article [Data quality: why it determines AI success](/blog/data-quality-for-ai-success/).

## Where AI fits into the digital twin

AI and the digital twin complement each other. The twin provides AI with time-aligned data in context, and AI gives the twin capabilities that physical models struggle to reach on their own.

- **Anomaly detection:** machine learning models detect deviations from normal behavior across many measurements at once.
- **Forecasting:** remaining useful life or scrap risk is predicted from historical patterns.
- **Surrogate models:** compute-intensive simulations are replaced by learned approximations that run in seconds and thus become usable in live operation.
- **Access through language:** language models make the twin's findings accessible to shift supervisors and maintenance teams.

Sequence is decisive. A forecast built on patchy data produces false alarms, and the plant team loses trust within a few weeks. The data pipeline comes first, the forecast second.

## Implementation in three stages

Implementation succeeds in three stages, each with its own acceptance. This limits risk and delivers measurable value early.

1. **Stage 1, decision and baseline:** one line, one decision, one metric. The failure analysis and business case are ready after six to eight weeks.
2. **Stage 2, digital shadow in live operation:** the data pipeline is in place, the model runs on live data and issues recommendations that people implement. After three to six months, the comparison with the baseline shows whether the effect materializes.
3. **Stage 3, feedback and roll-out:** proven recommendations are adopted as rules in planning or control. Further lines follow on the same data architecture.

These durations are guide values from delivery practice and extend when assets do not expose data.

**Three routes to a digital twin**
Criterion | Simulation project by engineering | Platform purchase with vendor project | With Niehues Consulting
Starting point | Technical model of the asset | Platform feature set | Decision with a baseline in euros
Data foundation | Design data, maintained manually | Integration per standard catalog | Data pipeline built to the decision's needs, validated with the plant
Business case | Rarely calculated | Presented by the vendor | Confirmed with real values before roll-out
Operation after the project | Open | License agreement, ownership open | Model ownership and metric from day one
Suited to | Planning and design | Groups with their own integration team | Investment decisions for plant and line

## Organization and operation

A digital twin is a product with a lifecycle. Every modification of the line and every new product variant changes the model. Two responsibilities should therefore be named before the start. Asset ownership in the plant decides whether recommendations are implemented. A model owner ensures the accuracy and upkeep of the model including its data pipeline.

Two legal questions belong on the list early. If the data relate to employees, for example log-ins at operator terminals, works council co-determination in Germany is engaged and the GDPR applies. For connected machines, the EU Data Act also governs who may access the data generated. This belongs in contracts with equipment suppliers.

## Niehues Consulting as your partner for the digital twin

[Niehues Consulting](/) supports digital twins from the decision through to operation: business analysis and business case, the sequence of lines, a prototype on real data and roll-out in the plant. Marin Niehues is a management consultant with a delivery perspective: he builds organizations that deliver and brings AI into operation. His projects range from VC-funded scale-ups to DAX 40 groups, and he has worked in 20+ countries across six continents.

The approach follows the Niehues Delivery Model with the phases Understand, Build and Implement. Three review questions matter most for the digital twin. What is the baseline, and how do we measure success? Does the prototype prove its value in the real workflow with real data? Who owns operation from day one? How such a project fits into AI implementation as a whole is described in the article [From pilot to scale](/blog/from-pilot-to-scale-ai-operations/). The starting point is a free 30-minute initial consultation via [the contact form](/#kontakt).

**First the decision, then the model**
A digital twin creates value when it improves a recurring, costly decision in the plant and that value is measured against a baseline. The data pipeline is the largest effort and the most important prerequisite. Starting with one line and a digital shadow produces a proven business case for roll-out within a few months.

**Sources**
- Bitkom: Digital twins in industry, representative survey of 552 manufacturers with at least 100 employees, 2025
- Bitkom: Digital twins in industry, representative survey of 555 manufacturers with at least 100 employees, 2026
- Kritzinger, Karner, Traar, Henjes, Sihn: Digital Twin in manufacturing: A categorical literature review and classification, IFAC-PapersOnLine, 2018
- National Academies of Sciences, Engineering, and Medicine: Foundational Research Gaps and Future Directions for Digital Twins, 2024
- ISO 23247-1 to -4: Digital twin framework for manufacturing, 2021
- IEC 63278-1: Asset Administration Shell for industrial applications, Part 1, 2023
- ISO/IEC 30173: Digital twin, Concepts and terminology, 2023
- Bundesverband Deutscher Unternehmensberatungen (BDU): fee study on management consulting, 2022

## Frequently asked questions

### What is the difference between a digital twin and a simulation?

A simulation runs on fixed assumptions and is started for a specific question. A digital twin is continuously connected to the real asset and synchronizes with its data. Simulations are often one component within a twin.

### How much does a digital twin cost?

Costs are driven mainly by scope and data availability, and the data pipeline is usually the largest single item. Model building, licenses and ongoing operation come on top. External consulting in the German market is typically billed at around 1,000 to 2,500 euros per consultant day.

### How long does it take to implement a digital twin?

As a guide value, a digital shadow for one line is in operation after four to seven months, including the business case and data pipeline. Rolling out to further lines is faster because the data architecture is in place. Older assets without a data interface extend the timeline.

### Is a digital twin worthwhile for mid-sized manufacturers?

A digital twin is worthwhile for mid-sized manufacturers when a bottleneck asset with high downtime or scrap costs is the focus. Starting with one line and a digital shadow keeps the investment small. A company-wide platform is not needed to begin.

### What is the Asset Administration Shell?

The Asset Administration Shell is a standardized digital representation of an asset, developed within the Industrie 4.0 initiative and standardized in IEC 63278-1. It describes the properties and services of a machine in a uniform way so that different software systems can read them. It is an important foundation for twins covering equipment from several manufacturers.

### Who is responsible for a digital twin in the company?

Business responsibility lies with asset or plant management, which decides whether recommendations are implemented. In addition, a named model owner is needed for accuracy and upkeep. Without this role, the model becomes outdated after the first modification of the line.
