Digital Manufacturing Platform: What Actually Has to Work

Last Tuesday I was in a customer's plant with their CIO, looking at the digital manufacturing platform he had bought eighteen months earlier. The dashboard on the wall was beautiful — colour-coded OEE per line, scrolling alarm feed, a 3D mock-up of the shopfloor rotating gently. Six of the eleven production lines in that plant were not on it. Not because the platform couldn't display them, but because nobody had ever managed to connect the machines on those six lines. The platform was real. The data was not.

This is the part of digital manufacturing platforms that the marketing pages skip, and the part that determines whether you have actually digitised your production or just installed a screen. After 35 years of connecting machines to whatever the prevailing platform of the decade happens to be — DDE in the 1990s, OPC Classic in the 2000s, OPC UA and MQTT today — I have a fairly settled view of what makes one of these platforms worth its license fee and what makes one of them an expensive screensaver.

What is a digital manufacturing platform, and what is it not?

A digital manufacturing platform is a software environment that sits above the production equipment and below the enterprise systems, and provides the connectivity, data processing, application surface and integration plumbing needed to run digital production use cases — OEE monitoring, traceability, alarm management, energy monitoring, AI-assisted operator support, and so on — without the customer having to wire each use case from scratch.

What it is not, in any honest definition: a magic box. A digital manufacturing platform does not turn a 1992 punch press into a smart machine on its own; somebody has to install a gateway. It does not replace MES, it usually contains MES functionality. It does not replace ERP, it integrates with ERP. And it is not the same thing as an industrial cloud — the cloud is where it lives, not what it is. The confusion of these terms in vendor marketing is one of the main reasons buyers end up disappointed.

What is actually inside a digital manufacturing platform?

If you take the marketing layer off and look at what's underneath, every digital manufacturing platform that works in the field — regardless of vendor — has the same six layers. The vendors that dress these up with proprietary names are not doing anything different; they're just renaming the same plumbing.

1. Physical connectivity. The gateways, taps and adapters that touch real machines. OPC UA / MQTT for modern equipment; digital I/O brownfield gateways for legacy equipment that has no digital interface; signal conditioners for analogue sensors. This is the layer the plant actually feels — installation time, plant disruption, retrofit cost — and it is the layer the demo rarely shows.
2. Edge processing. A small compute layer at the gateway or in a local industrial PC that timestamps, filters, buffers and pre-aggregates the raw signals before sending them anywhere. Without this, you either drown the network or lose data on the next outage.
3. Transport. The pipe that moves data from edge to cloud — typically MQTT or AMQP over TLS, sometimes OPC UA over the wire. Has to survive bad networks, intermittent GSM, plants where IT and OT don't talk to each other.
4. Semantic and execution core. The heart of the platform — the part that turns timestamped signals into contextualised production events, runs the MES logic (orders, batches, traceability, KPIs) and stores the historical record. This is where the differentiation between vendors is real, not cosmetic.
5. Application surface. Dashboards, mobile apps, shopfloor tablets, alarm interfaces, the SPC chart on the team-leader's screen. Important — but downstream of everything above. A perfect application layer on a broken capture layer is a kaleidoscope.
6. Integration adapters. The connectors to ERP (SAP, proAlpha, Microsoft Dynamics, Infor), to quality systems, to APS, to PLM. Often dismissed in vendor selection and then becomes 40 % of the project budget. Always ask for the list of pre-built adapters and what each one actually covers.

Five archetypes of digital manufacturing platform

Every vendor in this market — and there are now well over a hundred — fits into one of five archetypes. Understanding which archetype you're looking at is more useful than reading another feature comparison, because the archetype determines what the platform will actually be good at and what it will struggle with.

The first archetype is what we are at SYMESTIC. The other four are real choices for real customers; I have implemented projects against all of them and there are situations where each is the right answer. The mistake is buying an archetype based on the demo without knowing which category you're actually evaluating.

Where digital manufacturing platforms break in the field

From the projects I have either run, recovered or audited over the last decade, the failure modes cluster:

• The connectivity layer was undersold. The vendor demoed against a perfect OPC UA test machine. The customer's plant has 60 machines from 22 different manufacturers across four decades. Connecting them takes 6× the budgeted hours and stalls the whole rollout.
• The semantic layer is hardcoded. The platform models OEE, but only the way its developers thought OEE worked. Your plant calculates availability differently and the platform can't accommodate it without a custom development ticket. Six months later you're paying for changes that should have been parameters.
• The integration adapter was a slide. "We integrate with SAP" turns out to mean "we have done it once for one customer with their specific SAP module." Yours is different. Now it's a project.
• Nobody owns the data definitions. The platform is live, the data is flowing, and three people in three departments have three different definitions of "scrap." The dashboards disagree by design. The platform isn't broken; the data governance never happened.
• The OT/IT relationship was assumed, not built. Cloud platform requires outbound MQTT on the OT network. OT security policy says no. Project stops for nine weeks while a firewall ticket gets discussed. This one happens more than any of the technical failures.
• The vendor's customer-success team has 14 customers per person and you are number 12.

How to evaluate a digital manufacturing platform without getting fooled by the demo

The demo will look perfect. They always do. Here is the short list of things I push customers to ask, in order of how much they reveal:

1. "Show me your most heterogeneous customer plant." Not the cleanest one. The one with the worst machine mix, the oldest equipment, the most pre-existing systems. If the vendor doesn't have one, you are about to be one.
2. "Connect this brownfield machine in front of me." Bring a photograph of one of your real legacy machines (or, ideally, do this on a customer site visit). Ask how, exactly, the platform would connect it. Vague answers are the answer.
3. "What's the real go-live time for a 50-machine plant, including connectivity?" The marketing answer is "weeks." The honest answer for a real heterogeneous plant is two to six months depending on the connectivity work. Vendors who refuse to give you a real number are protecting a soft spot.
4. "Show me a customer reference whose plant looks like ours." Same industry, similar size, similar age of equipment. Different industries don't translate.
5. "Walk me through your last failed project." Every honest vendor has one. The ones who claim they don't are either lying or new.

Three questions I get asked on every project

"Do we need to replace our old machines first?"
Almost never. A 1985 lathe, a 1990 press, a 2003 CNC machining centre — every one of these can feed a modern digital manufacturing platform via the right gateway, with no PLC modification and no production interruption. The replacement-first mindset is the most expensive misconception in the market and I spend a lot of time arguing customers out of it.

"How is this different from just buying an MES?"
Most modern digital manufacturing platforms are an MES at their core, plus the connectivity, integration and application layers around it. Calling it a platform rather than an MES is partly accurate (it covers more than classical MES) and partly marketing (sounds more strategic). What matters is whether the MES functionality inside is real or shallow. Look at order management, traceability, KPI configurability — if those are thin, the "platform" is mostly a dashboard.

"Cloud or on-premise?"
For most of the mid-sized European manufacturers I work with in 2026, cloud is the right answer — faster deployment, automatic updates, lower IT burden, easier multi-plant rollout. The exceptions are real but narrow: heavily regulated environments where data residency is non-negotiable, plants with truly unreliable internet, customers whose corporate IT policy explicitly forbids cloud. For everyone else, the on-premise objection is usually a 2010 reflex applied to a 2026 problem.

If you are evaluating a digital manufacturing platform with a heterogeneous machine park — old equipment, multiple vendors, mixed protocols — that's the conversation I have most often. The SYMESTIC platform is built around the assumption that the connectivity layer is the hard part: brownfield IoT gateways for legacy equipment that has no digital interface, OPC UA / MQTT for modern equipment, an edge layer that survives bad networks, and a semantic core configurable enough that your definition of OEE doesn't have to match ours. Currently 15,000+ machines across 18 countries, average plant onboarded in weeks rather than quarters. If you want to compare your machine list against what we connect today, the fastest route is the Process Data overview and a 30-minute call.

Read next: MES — definition, functions and benefits · Data-driven manufacturing · Cloud MES vs. on-premise · Industrial IoT · OPC UA · Edge computing · OT/IT convergence · Smart factory · Industry 1.0 to 5.0 · Digital twin.

About the author

Martin Brandel

MES Consultant at SYMESTIC. 35+ years in industrial automation — Simatic S5 (1991), through S7, TIA Portal, OPC UA, MQTT and IoT gateways today. Former Head of Automation at SYMESTIC (2008–2018), MES & project lead since 2019. Specialist in brownfield connectivity and retrofit projects across heterogeneous machine parks. Dipl.-Ing. Communications Engineering. · LinkedIn