MES Manufacturing Execution System - Integration before function

What is an MES? — Definition and characteristics

A Manufacturing Execution System (MES) is an integration interface between operational technology (OT) and business IT (ERP). It is not primarily a feature product.

The MES sits at Level 3 in the ISA-95 hierarchy — exactly where sensor data from the shop floor is converted into business-relevant information for procurement, quality, and order processing. An MES does not simply collect production data: it orchestrates the data flow between manufacturing equipment (speaking protocols like OPC-UA or Fieldbus) and the ERP system (which needs production records, cost accounting, and supply chain data).

The core characteristics of an MES can be described as follows: real-time data capture from machines, personnel, and material flow — not exported to the ERP on a delay, but continuously synchronised. Integration bridge between operational technology and business IT — the MES translates machine language into business data. Quality and traceability orchestration — who produced what, when, where, and with what is documented. Resource optimisation logic — machine utilisation, personnel assignment, and material flow are coordinated in real time. Connection to regulatory requirements — documentation for GxP (pharma), TISAX (automotive suppliers), or ISO-9001 tracking. An MES is not a dashboard. It is also not a reporting layer. It is the nervous system between manufacturing and business.

Why an MES is now decisive for the DACH mid-market

Industry 4.0 is no longer optional — it is regulatorily relevant, competitively necessary, and the integration costs will force the decision on you.

The VDMA (German Engineering Federation) made it clear in 2024: the MES is the backbone of digital value creation in manufacturing. Not because it is a great tool, but because the data quality an MES delivers becomes the foundation for AI-driven anomaly detection, forecasting, and quality assurance. Bitkom reports that 42 % of German manufacturers now use AI in production — without an MES, you have no clean data to feed it.

The DACH mid-market is under pressure from two sides. On one hand, international customers (automotive, pharma) demand proof of digital traceability (TISAX, GDP, traceability). On the other, the EU AI Act draws boundaries from August 2026 onward: AI-driven anomaly detection in production falls into the "limited risk" class — anyone deploying it needs documented data quality. An MES without an explicit integration strategy cannot deliver it.

And the market signals are clear: the OPC-UA gateway infrastructure (needed to connect legacy machines to the MES) is growing 13 % annually in 2026. That is the signal: integration is the bottleneck, not feature breadth.

Case study: mechanical engineering firm with international project business

A kitchen manufacturer with ~2,000 employees had to introduce an MES — not because the existing production was poor, but because internationally coordinated project orders were no longer possible without MES-supported traceability.

The company had to deliver 1,500 kitchens for an international project — with a supply chain spanning three DACH countries, letter-of-credit documentation, and regulatory approval workflows. The reality: six different production managers within the company had developed six different order-to-cash processes. No one could centrally answer the question, "Where are the 1,500 kitchens right now?" — traceability ran on spreadsheets, email, and memory.

The MES (coupled to the ERP system) was not the first phase. The first phase was harmonising business processes. Who owns what, and when? Which data is critical? Which integration points are non-negotiable (because the authorities require them)? The MES itself was then configured, not built. The choice fell on a more cost-effective cloud solution — not because it had the most features, but because its integration API to the existing systems (ERP, quality database, CAM systems) was clearly defined. The project took 8 months, 1.5 technical FTE, and 0.5 FTE of process consulting. That was expensive. But the absence of integration chaos later made it worthwhile. The lesson: the MES was the solution to an integration problem, not a feature problem. Focusing on the "best features" would have led to the wrong system.

What most MES consultancies don't tell you

MES vendors and consultants often play the same role: they sell features. Three things that systematically get short shrift in the pitch.

1. Integration costs later what the license saves today

Across more than 1,200 ERP selection projects we have seen the pattern repeat: mid-market companies choose a cheap cloud MES with attractive dashboards, and 12–18 months later the real costs surface — in the integration with existing systems. An example: a mechanical engineering company chose an MES that spoke REST API, but their ERP system communicated via SFTP file import. The MES vendor said, "It integrates via the API!" — what they meant was, "We won't build it." The integration effort was 60 % of the original license decision. Fraunhofer IOSB has an approach for this: integration modelling comes before vendor selection. First: which data flows are critical? Then: which vendor can deliver those data flows? The second step is the cheaper one.

2. OPC-UA standardises integration — but only when it is configured

VDMA and ZVEI have preached it for a long time: OPC-UA is the standard for secure, vendor-neutral communication between legacy machines and modern MES. The reality in the DACH mid-market: 70 % of the machines on the shop floor are older than 10 years. They do not have OPC-UA natively. They need a gateway — and that costs money and requires specialist knowledge. This is not an MES problem. This is a legacy-integration problem. Most MES projects underestimate the complexity. The OPC-UA gateway market is growing 13 % annually, not because the technology is new, but because everyone is finally admitting it: legacy integration is the core effort.

3. Vendor licensing often hides in maintenance

A mid-market company pays X per month for the MES system. What's in it? Server license, user license, modules (quality, resources, reporting). What's not in it? Upgrades. Increasingly, "advanced AI anomaly detection" is a separate add-on. The vendor lures with feature breadth in the pitch — then comes the reality of the upgrade roadmap. Our reading: choosing an MES means making architectural decisions, not ticking feature boxes. The cheaper solution with a clear integration strategy beats the more expensive one with undefined integration ownership.

How we approach MES projects methodically

Integration-first methodology means this: before you evaluate a vendor, you define the integration interfaces.

Phase 1: integration modelling — What are the critical data flows? ERP → MES? MES → ERP? MES ↔ SCADA? Clarify this first — not as a list, but as an architecture diagram with ISA-95 levels as the reference. Phase 2: legacy inventory — Which old machines speak which protocols? The reality on the floor is the basis for realistic budgets. Phase 3: vendor shortlist by integration fit — Not "which has the best features?", but "which can deliver our critical data flows, with APIs for our legacy systems, without custom code?" Phase 4: proof-of-concept with real data flows — A weekend's effort, small cost, but it exposes the real integration risks. Phase 5: go-live with integration ownership — Define clearly: who is responsible when data sits in the MES and does not reach the ERP? This must be in writing before day one of implementation.

Common mistakes in MES projects

Three mistakes we have seen repeat across more than 30 years of project work.

Mistake 1: feature-first vendor selection. The system has the prettiest dashboards in the demo. Nobody asks, "How does this integrate with our SAP?" Then go-live arrives, and the integration reality is more expensive than the system itself. The solution: an integration checklist before the demo.

Mistake 2: cloud migration too fast. A mid-market company says, "Cloud is modern, we'll go cloud." But their ERP still runs on-premises behind a TISAX-certified firewall. The cloud MES can't simply reach in from outside. Now you need a VPN, a hybrid-architecture design. The solution: cloud vs. on-premises is an IT-architecture decision, not a trend decision.

Mistake 3: wrong project governance. The MES is treated as an IT project. But it is a business-process project. The IT lead says, "We need a database with a REST interface." The production managers say, "We need real-time transparency." Those are two different requirements. The solution: a steering committee made up of business, IT, and operations from day one.

Frequently Asked Questions

How is MES different from ERP?

Fundamentally. An ERP is a business data hub — it stores orders, invoices, cost accounting. An MES is a real-time loop — it actively steers production flow, synchronises machine data, and feeds filtered, structured data back to the ERP. They overlap in quality function but with different responsibilities. The ERP says: "This batch failed conformance — initiate recall." The MES says: "This machine exceeded temperature by 2 degrees — real-time alarm."

Does our DACH Mittelstand company really need a standalone MES, or would an ERP-MES module suffice?

It depends on manufacturing complexity. A simple-serial producer (say, frame manufacturer with five machines, one variant) could manage with a good ERP-MES module (SAP ME, Oracle MES). A company with project business, international supply chain, or regulatory requirements (pharma, automotive suppliers) needs a real, separate MES. The rule of thumb: if production logic is more complex than "material in, product out, measure quality," then a real MES is the right play.

Which MES system does Dreher Consulting recommend?

We don't recommend a specific system — that would be vendor lobbying. Instead: follow the integration-modelling process. Once your critical data flows are clear, the right systems become visible. Sometimes it's a large ERP vendor (SAP, Oracle). Sometimes it's a specialist MES (Plex, Parsec, GlobalShop Solutions). The best choice is always the one that serves your integration requirements without custom code.

How long does MES implementation take?

That's driven by your legacy-integration complexity. A greenfield production site (new machines, clear data flows, modern ERP): 4–6 months. A brownfield site (old machines, multiple legacy systems, complex orders): 9–15 months. The timeline hinges on Phase 2 (legacy inventory): the uglier the reality there, the longer the project. That's not a feature; that's reality.

Next steps

If you want to go deeper on these topics, or similar questions about MES integration are surfacing in your company, the team at Dreher Consulting is your point of contact. We offer a no-commitment 30-minute conversation — in which we understand your specific integration situation and work out concrete next steps with you.

Dr. Harald Dreher
Owner & Senior Consultant · Dreher Consulting ^®

Dr. Harald Dreher has advised mid-market managing directors in Germany, Austria, and Switzerland (DACH region) on digitalisation, ERP, and AI strategy decisions for more than 30 years. Over 1,200 completed projects. Owner-led, vendor-neutral, with proprietary AI model SCOReX®.

About Dr. Harald Dreher · Editorial standards

Manufacturing Execution System (MES): Integration Before Functionality