|
→ White Paper “Management of Quality-Relevant Production Data”: How to consistently capture, audit-proof archive, and efficiently analyze process data—without siloed solutions. |
KEY POINTS AT A GLANCE
|
IN A NUTSHELL
|
The term “process management” is used in various ways. In the administrative world, it refers to the modeling of business processes. In manufacturing, process management software refers to something more specific: the recording, monitoring, and documentation of physical production processes based on their measurement data.
At its core, it consists of four functional blocks. First, data acquisition: The software receives data packets from plant control systems, checks them for accuracy, and stores them in a structured format. Second, monitoring: Threshold values are automatically checked, and any deviations trigger a notification. Third, evaluation: Key metrics such as cp, cpk, or Sigma are calculated and presented graphically. Fourth, documentation: A history file is created for each component, which remains verifiable even years later.
Real-time capability is crucial. Software that only shows deviations in a daily report does not prevent defective parts—it merely counts them. Only when an exceeded threshold triggers an alert within seconds can the operator or process manager intervene before the next batch is affected.
A concrete real-world example from the automotive industry: At the Mercedes-Benz plant in Hamburg, more than 14 axle assembly lines are connected to a process data management system. The most noticeable effect was not a single metric, but rather the elimination of manual data collection. The time previously spent gathering data from various sources is now saved because the process data is available centrally and updated daily.
Most mistakes made when selecting process management software occur because people focus on feature lists rather than operational suitability. A long list of features says little about whether the software will work with a company’s existing, historically evolved plant infrastructure.
By far the most important criterion is the vendor-neutrality of the interfaces. The average plant operates equipment from multiple manufacturers spanning several decades. Software that connects only to equipment from a single manufacturer does not solve the problem of siloed systems—it merely shifts it. Only when older controllers, ASCII telegrams, and modern protocols are processed equally does the end-to-end view emerge—which is what process management is really all about.
The following overview ranks the key criteria according to their importance for project success.
| Criterion | What It’s About | Why it matters |
|---|---|---|
| Vendor-neutral interfaces | Integration of systems and control units from different manufacturers and of different vintages | Without broad connectivity, new siloed solutions emerge instead of a consistent view of data |
| Real-time monitoring | Threshold checks and alerts within seconds | This is the only way to prevent defective parts rather than just counting them |
| Traceability | Complete history records for each component, in some cases tracing back to the supplier | The basis for quality documentation and liability issues |
| Scalability and modularity | Gradual expansion without changing systems | Protects the investment and avoids a risky “big bang” |
| Integrated archiving | Audit-compliant, long-term storage of process data | Meets retention requirements and keeps production databases lean |
Archiving is an aspect that is often underestimated. Depending on the industry, process data must remain traceable for many years. If this data remains in the active production database, it grows unchecked and analyses become slower. Well-designed process management software separates active data from archived data without losing access to the latter.
In regulated industries, process management is not a matter of efficiency, but of the obligation to provide evidence. Manufacturers in the automotive, medical technology, or aviation sectors must be able to demonstrate that processes operate within defined limits.
The key standard in the automotive industry is IATF 16949. Its requirements for documented information (Section 7.5), production control (Section 8.5.1), and, in particular, traceability (Section 8.5.2) can hardly be reliably met without systematic process data collection. In addition, ISO 9001:2015 requires data-driven decisions in Section 9.1 and risk-based thinking in Section 6.1. Both of these require that process data be available in a form that can be analyzed.
Added to this is the stricter EU Product Liability Directive of 2024, which broadens the definition of “manufacturer” and alters the burden of proof in cases of product defects. For manufacturers, this means that comprehensive process documentation is shifting from an efficiency advantage to a protective measure.
An important caveat concerns AI-supported analyses. Modern process management software increasingly incorporates AI, for example, to detect anomalies in process curves. Under the EU AI Act, high-risk systems are subject to specific requirements regarding transparency and human oversight. In safety-critical industries, AI must never make fully autonomous approval decisions. It provides decision support, but the final responsibility remains with humans.
The implementation of process management software rarely fails because of the software itself, but rather due to unrealistic planning. A “big bang” rollout across all facilities at once unnecessarily increases the risk. A phased approach has proven to be effective.
The following steps describe a typical implementation path that has proven effective in projects.
When making a selection, honesty about limitations is more important than any marketing claim. Process management software is a specialized tool, not a jack-of-all-trades.
First, it does not replace a full-fledged MES. A Manufacturing Execution System (MES) controls production orders, resources, and capacities across the board. Process management software is closely related to an MES but focuses on the quality-related aspects of processes and data. Anyone looking for a complete MES should clarify this distinction from the outset.
Second, it does not solve data quality problems at their root. If sensors are incorrectly calibrated or characteristics have been defined imprecisely, the software will reliably document the erroneous values. Ensuring data quality at the source remains the responsibility of the production department.
Third, it does not make autonomous decisions. Even with AI support, the approval of a component in safety-critical industries remains a human responsibility. This is not only a technical limitation but also a regulatory one.
These limitations are not an argument against using the software—quite the contrary. Those who are aware of them can plan for the software realistically and avoid disappointments arising from unrealistic expectations. For a more in-depth look at the underlying data parameters, see the article on “Collecting Process Data Without Data Chaos.”
Manufacturing process management software records, monitors, and documents measurement data from physical production processes such as screwing, press-fitting, or bonding. It consolidates data from various systems, automatically checks threshold values, and issues real-time alerts in the event of deviations. The goal is to provide a consistent, verifiable view of process quality. This distinguishes it from administrative process management, which models business processes.
An MES controls the entire manufacturing execution—that is, orders, resources, and capacities. Process management software focuses on the quality-related collection and evaluation of process data and is therefore more specialized. It is closely related to an MES but does not replace a full-fledged MES. Many companies use both systems in parallel and integrate them via interfaces.
The most important criterion is that the interfaces be vendor-neutral. Since plants typically operate equipment from different manufacturers and of varying ages, the software must be able to connect to both older and modern control systems equally. Software that supports only select systems creates new silos. Only broad connectivity provides the end-to-end data view that delivers the actual benefit.
In the automotive industry, IATF 16949 is central, particularly the requirements for documented information (Section 7.5), production control (Section 8.5.1), and traceability (Section 8.5.2). In addition, ISO 9001:2015 requires data-driven decision-making (Section 9.1) and risk-based thinking (Section 6.1). In addition, the EU Product Liability Directive 2024 tightens the burden of proof in cases of product defects. For AI-supported analyses, the transparency and oversight requirements of the EU AI Act also apply.
In safety-critical industries, fully autonomous approval decisions made by software are not permitted by regulation. The software provides decision support by highlighting deviations and providing key metrics. The final approval remains the responsibility of a human. This also applies to AI-supported functions that, under the EU AI Act, require human oversight.
The duration depends on the number and diversity of the systems to be integrated. A phased approach has proven effective, starting with a pilot system where interfaces and data quality are tested before additional systems are integrated. A “big bang” approach—implementing the system across all plants simultaneously—unnecessarily increases the risk. In practice, companies connect plants successively over the course of weeks and months until the entire plant is covered.