Digitalization in Production: What Actually Works

What is digitalization in production?

Digitalization in production is the replacement of paper-, estimate- and memory-based manufacturing operations with continuous, machine-readable data that flows in real time between the shopfloor, the planning layer and the business. In a properly digitalized plant, every cycle, every stop, every order, every reject and every process parameter exists as an event with a timestamp, attached to a product, an order and a machine, and accessible to anyone who needs it within seconds of it happening.

I have been working on this problem since 1989, which is long enough to have seen four distinct waves of it pass through the industry. Each wave arrived with the same marketing posture: this time everything changes. Two of the four actually moved the needle. Two did not, at least not for the mid-market manufacturer. This article is my honest attempt to tell the difference, because the mid-market plant asking "where do we start with digitalization in 2026?" is getting very bad advice from almost everyone who writes about this topic.

Four waves, 35 years: what actually delivered

Each of the waves below promised to transform manufacturing. In the view from my desk — having personally led or observed hundreds of plants across each of them — only two delivered the transformation their marketing promised. Recognising which is which is the starting point for every sensible digitalization decision a mid-market CEO or COO makes today.

The pattern jumps out once you write it down. The waves that delivered (process control, cloud-native MES) lowered the cost of entry. The waves that disappointed (classic MES, Industry 4.0) raised it. In a fragmented industry where 80 percent of plants are below 1,000 employees and cannot absorb a €1M digital transformation project, that single criterion — does this wave lower or raise the cost of participation — predicts almost everything.

What to digitalize first, second, third

Mid-market CEOs routinely ask me where they should start their digitalization programme in 2026. The conference-stage answer involves digital twins, AI copilots and autonomous scheduling. The honest answer is far less glamorous and works every single time. Do these three in this order. Skipping is where it all goes wrong.

First: digitalize the basics of what is actually happening on the floor. Real-time capture of cycles, stops, rejects, orders, setup events and operator actions at 1-second resolution, tied to the production order. This is MDE (machine data acquisition) and BDE (operational data acquisition) done properly. Without it, everything above it is fiction. With it, 80 percent of the value that mid-market digitalization is supposed to deliver is already on the table. I have said this same sentence to customers for three decades and it is still true.

Second: digitalize the loop between the data and the decision. OEE on screen at the machine, not in a Friday Excel. Alarm events categorised automatically, not after the fact. Quality rejects linked to the process parameters that caused them. ERP actuals flowing back from the floor without the shift leader typing them in. This is where the 5-15 percent improvement numbers every MES vendor quotes actually get realised. The numbers are real — I have watched them happen — but only when the first layer is honest.

Third, and only then: the modern additions. Predictive maintenance, digital twins, AI-assisted scheduling, energy optimisation, mobile shopfloor apps. Every single one of these becomes genuinely valuable once layers one and two exist. Without them, every one is a demo. The plants that are running successful AI projects in 2026 almost all spent 2019-2023 getting their data layer right. The plants that are still struggling are the ones that tried to start at the top.

The first-measurement shock: in 25 years of switching on real-time data capture in customer plants for the first time, almost every single plant has discovered that reality is 15-30 percent worse than they believed. Availability is lower, microstops are more frequent, setup times are longer. Not because the staff work badly, but because without data, perception is systematically distorted. Every real digitalization programme passes through this moment. The plants that handle it well use it as the starting point for improvement. The plants that hide it have lost before they began.

Why most mid-market programmes stall

The industry has been writing "digitalization roadmaps" for fifteen years, and the failure rate has barely budged. I see the same five patterns repeat at plant after plant. Naming them explicitly is usually enough to stop them.

1. Strategy before data. A plant hires a consultancy to produce a 60-slide digital strategy. The deck is beautiful. Six months in, nobody has connected the first machine. Twelve months in, the budget is half-spent and the CEO wonders what happened. The right sequence is the reverse: connect one machine, see real data, let the strategy emerge from what the data reveals.

2. Waiting for the perfect machine park. "We want to digitalize, but first we need to replace these old presses from 1998." This is a trap. Old machines can almost always be connected without a PLC change, without production interruption, through digital-I/O gateways or retrofit sensors. I have yet to encounter the machine that genuinely cannot give up its cycle count. The wait is almost always a procurement habit, not a technical limit.

3. IT-led instead of production-led. IT picks the platform based on enterprise architecture alignment. Production, who will actually use it, is informed at month nine. Adoption never happens. The working pattern is the reverse: production defines the use cases, IT ensures the security and integration constraints are met. Both sides matter. Only one of them should drive.

4. Top-down rollout across all plants at once. Corporate mandates a digitalization initiative across six plants simultaneously, with central standards set before any site has proven the approach. The result is uniform failure, on time and on budget. The working pattern: prove it at one plant in 8-12 weeks, let the second plant copy the working version, let the rest follow. This is how the Meleghy rollout scaled to six sites in six months and how the Carcoustics rollout reached 500+ machines across seven countries in under a year.

5. Treating digitalization as a project with an end date. A plant finishes "the digitalization project" and moves on. Two years later the system is half-used and the data is stale. Digitalization is an operating model change, not a project. The plants that succeed assign a permanent owner, integrate the data into daily management, and keep extending the scope year after year.

The pattern that works, in one sentence

Start with one line, one workshop, one real data point. Get it on a screen that same day. Let the operators and the shift leader see what the data says. Let the gap between what they believed and what is actually happening become the starting point for every improvement conversation that follows. Everything else — ERP integration, KPI dashboards, additional plants, AI on top — grows out of that first honest data point. Plants that get this right measurably outperform plants that start with a 60-slide strategy. I have been watching this pattern play out for 35 years and it has not changed once.

A real case: Schmiedetechnik Plettenberg

Schmiedetechnik Plettenberg is a mid-sized metal-processing specialist producing forged components in a discrete manufacturing environment with highly variable batch sizes, demanding setup procedures and parallel machine chains. The starting point was the classic mid-market digitalization problem: InforCOM ERP handled the order world well, production data was captured mostly on paper, machine states were only partially visible, and deviations were typically recognised after the fact rather than during the shift.

What makes the Schmiedetechnik case worth telling is not the result but the sequence. The engagement did not start with a digital strategy. It started in the factory. In a hands-on workshop, the first machine was connected, data points were defined, dashboards were configured live. Within hours the team at the line saw real-time cycle times, quantities, stops and process deviations — most of which were different from what had been assumed.

From that first data point, three things grew in sequence. First, the InforCOM integration: once a production order is released in the ERP, all relevant work steps, machine information and time data are automatically available in the SYMESTIC layer, and all feedback (quantities, times, stops, status) flows back to the ERP without manual intermediate steps. Second, the enablement workshop: key users were trained to connect additional machines themselves, adjust dashboards, and configure new use cases — because digitalization that only the vendor can extend is not digitalization, it is dependency. Third, the expansion into adjacent use cases, driven by the team on the floor, not by an outside roadmap.

The measurable results speak for themselves, but the cultural result matters more: real-time transparency across machines, shifts and orders; faster root-cause analysis; fewer manual entries; a clean data foundation that the next AI project can actually use. "SYMESTIC gives us continuous real-time transparency we did not have in this form before," said Thorsten Manns, Technical Director at Schmiedetechnik Plettenberg. That is the voice of a plant that has genuinely digitalized, not a plant that is planning to.

FAQ

Is digitalization in production the same as Industry 4.0?
No. Digitalization is the practical, often unglamorous work of replacing paper and estimates with continuous data. Industry 4.0 is a marketing and policy framework that tried to sell the entire end-state (smart factories, digital twins, autonomous production) in one bundle. A plant can be very well digitalized without ever using the phrase Industry 4.0. The reverse, a plant that talks about Industry 4.0 but cannot show its current OEE in real time, is the more common pattern.

Where should a mid-market plant start in 2026?
Connect one machine this week. Not a strategy, not an RFP, not a steering committee. One machine, one gateway, data visible on a dashboard within hours. Everything else follows from that. Plants that start at the bottom reliably outperform plants that start at the top, at a fraction of the cost and time.

Do we need to replace our old machines to digitalize?
No, and the belief that you do is the single most common reason digitalization stalls. Machines from the 1990s can almost always be connected through digital-I/O gateways that tap existing signals (cycle pulse, running contact) without any PLC change, without production interruption. The Klocke packaging plant was connected entirely this way across an entire site. The wait for "better machines" is almost always a procurement habit, not a technical limit.

How long does a digitalization programme take to show real results?
First real-time data visible on the shopfloor: hours to days, not months. First measurable improvement in availability, OEE or scrap: typically 8-12 weeks. The Meleghy rollout across six plants in Europe delivered 10 percent less downtime, 7 percent more output and 5 percent higher availability within six months. The Neoperl programme delivered 8-15 percent on multiple KPIs within a year. These are measured, not projected, numbers.

Cloud or on-premise for a mid-market digitalization programme?
In 2026, cloud for almost every mid-market case. The arguments that justified on-premise in 2015 — data sovereignty, integration complexity, latency — have largely been solved by hardened gateways, mature cloud providers and standards like OPC UA and MQTT. Cloud-native MES is the first wave of this entire topic that genuinely works for a plant with limited IT capacity, which is most of the mid-market. On-premise still makes sense in heavily regulated contexts and in the largest enterprise groups with specific compliance constraints.

What role does AI play in production digitalization right now?
Real but overhyped. AI on top of a plant without a proper data layer is a demo, nothing more. AI on top of a plant with honest, real-time, order-linked data is where genuine advantage is emerging in 2026 — pattern detection in alarm data, correlation of process parameters with defects, shift optimisation. The plants that will benefit most from AI over the next three years are not the ones rushing to it; they are the ones who spent 2020-2024 quietly getting their data right.

How does SYMESTIC support production digitalization?
The SYMESTIC cloud-native MES platform is designed exactly for the mid-market digitalization path described in this article: go-live in days through IoT gateways (OPC UA, digital I/O, MQTT) that connect any generation of machine, real-time OEE and loss analysis out of the box, bidirectional ERP integration with SAP, Infor, Microsoft and proAlpha, and a modular catalogue that customers extend themselves. Over 15,000 machines connected, 18 countries, zero customer churn in 2024. See SYMESTIC Production Metrics.

Related: MES · OEE · Cloud MES · Production Data · Industry 4.0 · Smart Factory · IIoT · SYMESTIC Production Metrics

About the author

Uwe Kobbert

Founder and CEO of SYMESTIC GmbH. 35+ years in manufacturing — consultant at SAS (1989), industry lead at STERIA for process control and MES in food and beverage (1992-1995), founded SYMESTIC in Dossenheim in 1995. Led the cloud-native reinvention of the SYMESTIC platform in the mid-2010s. Today: 15,000+ connected machines, 18 countries, four continents, zero customer churn in 2024, financed without external investors. Dipl.-Ing. Nachrichtentechnik/Elektronik. Nominated for the Großer Preis des Mittelstandes. · LinkedIn