MES, ISA-95 and ERP/PLC integration in pharma: a pragmatic architecture for Swiss sites

The MES discussion comes up at every architecture review on Swiss pharma sites. The plant director wants end-to-end electronic traceability. Quality wants a bulletproof 21 CFR Part 11 audit trail. IT wants a single source of truth. Automation just wants the line to run. And the MES vendor's sales rep comes in to explain that their product handles all of it on its own.

It doesn't. No MES covers everything, and the one that tries to reach 100 % becomes unmanageable within three years. This article proposes a pragmatic grid for Swiss pharma sites, including teams running pharma automation integration in Vaud, Valais and Geneva who must recommend an architecture that holds up under a Swissmedic, FDA or EMA audit.

Our angle is not the vendor's. It's the angle of an automation engineer who spends days between a Siemens TIA Portal or Rockwell Studio 5000 PLC layer, a SCADA, a historian, and an MES that must receive its manufacturing orders from a SAP or Oracle ERP. Five years of regulated pharma work taught us one thing: the architecture that holds up over time is the one that respects ISA-95 boundaries.

ISA-95: the four-level reading grid

ISA-95 (ANSI/ISA-95 and IEC 62264) is not a cookbook. It is a reference model that splits the plant into logical levels. Many architecture debates die down as soon as stakeholders agree to name the levels and respect the borders.

Level 0: the physical process. Sensors, actuators, calibrated instruments. Whatever touches the product.
Level 1: direct control. Programmable logic controllers (PLC), drives, PID loops, safety systems. Hard real time, cycles in the millisecond range.
Level 2: supervision. SCADA, HMI, DCS. Operator visualization, alarm management, short-term historization. Soft real time, latency around a second.
Level 3: manufacturing operations. MES, LIMS, recipe management, batch tracking, production dispatching, in-process quality. Time horizon: minute to day.
Level 4: enterprise planning. ERP (SAP, Oracle, Microsoft Dynamics), order management, site and corporate inventory, analytical accounting. Horizon: week to year.

The mapping is useful in itself, but the real value lies in the interfaces between levels. That's where projects derail.

The critical interface on a pharma site is Level 3 to Level 1-2: how the MES drops a manufacturing order, how it reads back the executed process parameters, how it seals an electronic batch record. The second critical interface is Level 4 to Level 3: how the ERP transmits a production plan to the MES without the MES becoming the receptacle for the ERP's business complexity.

Market MES products and their DNA

There are about twenty MES vendors in pharma. Three dominate in European life sciences: Werum PAS-X (Körber), Siemens Opcenter Execution Pharma (the former Camstar refactored after the Apriso merger), and Rockwell PharmaSuite. A handful of others matter in niches: Emerson Syncade, Honeywell Workcenter, Tulip for guided manual operations, and the legacy in-house developments still alive on historic sites.

Each has a DNA, and that DNA explains the positioning better than any brochure.

Werum PAS-X

PAS-X was born in the German pharma industry and was acquired by Körber. Its historical strength is the configurable eBR (electronic Batch Record) and the maturity of its GMP coverage. Its natural perimeter is broad: recipe management, dispatching, eBR, deviation management, equipment and material handling. It integrates by default with SAP ERP (large SAP pharma installed base).

The pitfall: PAS-X is heavy to configure. A nominal deployment takes 12 to 24 months. The cost of configuring electronic recipes explodes if you cannot factor the variants. And the boundary with Level 2 (SCADA) remains to be drawn: PAS-X can drive process steps directly, or delegate them to the SCADA. That choice is structural and is rarely formalized in early specifications.

Siemens Opcenter Execution Pharma

Opcenter (formerly Camstar Pharma, then merged with Apriso for manufacturing operations management) benefits from native integration with the Siemens stack: PCS 7, WinCC, TIA Portal, Opcenter APS for planning. For a site already aligned on Siemens at Level 1-2, the integration argument is serious.

The pitfall: Opcenter Pharma has been a product in consolidation since the Camstar acquisition in 2014, and the roadmap has gone through several inflections. For a 2026 project, check very carefully the target version, the modules available in that version, and the 7-to-10 year support commitment. The entry ticket is competitive, but total cost of ownership depends heavily on the functional scope retained.

Rockwell PharmaSuite

PharmaSuite is Rockwell's pharma offer, based on the FactoryTalk platform. It is typically the choice on a site with a Rockwell ControlLogix and FactoryTalk View automation history. Vertical integration down to the PLC is smooth, and the philosophy is more modular than PAS-X.

The pitfall: on the European pharma market, the PharmaSuite integrator ecosystem is thinner than for PAS-X or Opcenter. For a Swiss site, identifying a competent partner who will survive the project is a topic in itself. And the fine-grain eBR coverage is less mature than PAS-X.

Other options worth knowing

For small sites or tactical capacity extensions, two families deserve a look. First, light MES platforms like Tulip or Rootstock, which address guided manual operations and simple eBR without the weight of PAS-X. Then custom solutions on modern stacks (Mendix, OutSystems, even Node.js + Postgres validated GAMP5 category 5), which remain defensible for very business-specific perimeters but demand exceptional validation discipline.

Custom GAMP5 category 5 in pharma is viable. It carries a maintenance and re-validation cost on every change that is not trivial. Do not pick it unless you have the in-house team to carry it through the years.

The « MES that wants to do everything » trap

The most expensive mistake we see in the field is not the wrong MES choice. It is the right MES, used to do the SCADA's or the ERP's job.

Three symptoms ring the alarm:

The MES writes process setpoints directly into the PLC. If your MES sends PID setpoints straight to a controller, you have probably bypassed Level 2. At the next MES upgrade, process recipes head back to OQ revalidation. That's heavy and avoidable.
The MES manages customer orders or raw material BOMs. That's the ERP's role. If the MES takes it on, you have two sources of truth that will eventually diverge.
The MES manages site Windows access audit trail. That's Active Directory's role. The MES should rely on it, not rebuild it.

The pragmatic rule: each function should live at the lowest ISA-95 level that can deliver it. Anything that can stay at Level 2 (SCADA, historian) stays at Level 2. Anything purely business stays at Level 4. The MES handles manufacturing execution and regulatory traceability, not the rest.

This discipline makes revalidations simpler. When the ERP migrates from SAP ECC to S/4HANA, the MES impact must be containable to the Level 4 interface. When the SCADA changes version, the MES impact must be containable to the Level 2 interface. If the architecture is correct, these migrations are plannable. If it is tangled, every change is a transformation project.

A concrete case: a Swiss fill & finish site that re-scoped its MES

We recently worked with a sterile filling site in French-speaking Switzerland that had inherited a PAS-X configured over ten years, with a functional perimeter that had become impossible to maintain. The MES was handling manual weighing in the preparation room, batch dispatching, the filling line eBR, parametric release, and a deviation module that overlapped with the site QMS.

The rationalization project did not change vendors. It re-scoped:

The deviation module was moved to the official group QMS. Duplicate removed.
Manual weighing was migrated to a dedicated Tulip application, validated GAMP5 category 4, with CSV interface to PAS-X. Validation cost on that perimeter divided by three.
The filling line eBR stayed in PAS-X. That is its core business.
Parametric release was redesigned to rely on rules encoded in the LIMS, with PAS-X aggregating results only.

The project ran eight months, including three months of scoping with quality, production, IT and automation teams. The benefits are measurable: shorter eBR cycle time, IT/OT load reduced, process deviations documented without manual copy-paste between PAS-X and the QMS.

The key point: no new structural software. Just a re-scoping of the functional perimeter, level by level, respecting ISA-95.

Scoping MES at specification stage

When you write the specification for a new MES (or a perimeter refresh), three questions save entire quarters of project time.

First question: which business processes actually need an MES, and which can live at Level 2 or 4?

Walk through the full process list: weighing, dispatching, recipe execution, in-process controls, parametric release, deviation management, change control, operator training, equipment management (maintenance, calibration). For each, ask: does this really need the MES layer, or does another system already handle it well?

Second question: which ISA-95 interfaces will I build, and on which protocols?

ERP to MES (Level 4-3): typically SAP IDoc or REST API. MES to SCADA (Level 3-2): OPC UA, MQTT for process data flows, ISA-88 messaging (S88 unit recipes) for manufacturing orders. MES to LIMS: standard LIMS interface or dedicated middleware. Every interface should have an owner, a format, a protocol, and a defined SLA.

For Level 3-2 interfaces, the OPC UA question deserves its own architectural analysis. On fleets of several PLCs, the Pub/Sub vs Client/Server trade-off is not trivial, and we dedicate a separate article to it.

Third question: who validates what, and under which GAMP5 chapter?

The MES is typically GAMP5 category 4 (configured software) with a few category 5 custom modules. Interfaces must be validated point by point. The SCADA is generally category 4. PLCs hold category 5 custom code. The validation strategy must be set before kickoff, not discovered in OQ phase.

Our [CQV pharma feedback](/blog/cqv-pharma-best-practices) details the classic qualification pitfalls on this multi-level architecture, and our article on the [IT/OT and data bottleneck](/blog/automation-it-data-goulot-etranglement) goes deeper into interface questions between Levels 2 and 3.

What a pharma automation integrator should refuse

Five common requests that should trigger an architecture discussion, not silent execution:

« Put the MES in direct control of the PLC to save time ». Answer: no. Control through Level 2; the MES sends a recipe, not a setpoint.
« The MES has to manage suppliers and raw material orders ». Answer: no, that is the ERP. The MES consumes material data; it does not maintain it.
« We validate the MES once and never touch it again ». Answer: false. The MES is alive. You need a real Change Control process, not a myth of final validation.
« Slap an MES on it for the eBR, we will handle the rest later ». Answer: not before defining the Level 2 interface and the exhaustive perimeter. Otherwise you build architectural debt.
« Vendor X's MES does everything, so we take it everywhere ». Answer: no. A homogeneous multi-site MES architecture without per-site scoping leads to divergent configuration and integration debt that costs more than the freedom to choose locally.

Conclusion

MES architecture in pharma is not a product topic. It is a question of functional partitioning, level by level, that must hold up against an audit, an ERP migration, a SCADA refresh and a revalidation. The right reflex before picking a vendor is to lay down the ISA-95 map, place each process, and defend the boundaries.

If you are scoping an MES project on a Swiss pharma site, our [pharma sector page](/secteurs/pharma) and our [architecture service page](/services/architecture) describe the kind of scoping missions we run on this type of topic. And if you want to discuss a concrete case, [contact us](/contact) for a pre-vendor architecture review session.

FAQ

Is ISA-95 mandatory for a pharma site?

No, ISA-95 is not a regulatory requirement. It is a reference model that helps structure architecture. No Swissmedic or FDA inspector will ask for ISA-95 compliance. However, an architecture that respects this partition is easier to defend in audit, simpler to maintain, and more stable over time.

Do I need an MES for a small filling site?

Not necessarily. If your site manufactures few SKUs with stable recipes and modest volume, a paper eBR or a lightweight Tulip-style solution may suffice for several years. Moving to a structured MES is justified when traceability complexity, variant count or ERP integration requirements exceed what a team can handle manually.

What is the typical deployment cost of a pharma MES?

Depending on vendor, perimeter and site size, expect between 800,000 CHF for a minimal mono-line deployment to several million CHF for a multi-site rollout with full eBR. Licensing is often under 30 % of total cost. The rest is integration, configuration, validation and change management.

How do I choose between Werum PAS-X, Siemens Opcenter and Rockwell PharmaSuite?

The decision hinges on three factors: the existing automation stack (Siemens, Rockwell or mixed), the availability of a competent local integrator, and consistency with corporate group choices if you are a subsidiary site. For a greenfield independent Swiss site, PAS-X remains the safe pick, but the entry ticket is high. Opcenter is competitive in a dominantly Siemens environment. PharmaSuite makes sense on a historically Rockwell site.

What is the concrete difference between Level 2 (SCADA) and Level 3 (MES) in pharma?

The SCADA handles process in real time: operator visualization, alarms, historization, PLC sequence execution. The MES handles batch manufacturing: electronic recipe, material traceability, eBR, 21 CFR Part 11 electronic signatures, result transmission to LIMS and ERP. The SCADA is equipment-centric. The MES is product- and batch-centric. On a well-architected site, they talk via OPC UA and ISA-88, without overlapping.