The impact of Industry 4.0 on industrial productivity

Most Portuguese factories that talk about Industry 4.0 start with hardware — sensors, robots, touchscreens on the line. It's the wrong path. The real problem isn't a lack of equipment: it's a lack of visibility. Without real-time production data integrated into the ERP, the operations manager makes decisions with yesterday's information — and no sensor solves that if the information chain remains broken in the middle.

Portugal is the world's 3rd largest producer of moulds, with around 472 companies and production of ~€947M in 2023, of which 80% was exported to 86 countries (source: CEFAMOL, 2023). These companies compete with German, Czech and Chinese suppliers on deadlines and quality. International buyers are beginning to demand OEE reports and batch traceability as a condition of approval. The margin for invisible inefficiency is zero — and the same applies to any factory making metal components, plastic injection or garments that exports under someone else's brand.

This guide doesn't sell Industry 4.0. It gives you a method to audit where your factory stands today and what to do next — ready to use in tomorrow's meeting.

The sensor nobody reads changes nothing. The data that reaches the manager in under two minutes changes everything.

The thesis suppliers won't tell you

Industry 4.0 isn't a technology project. It's an operational visibility project. The difference matters because it radically changes what you buy first.

A metal components factory with 60 employees in the Aveiro–Marinha Grande corridor may have state-of-the-art CNC lathes and still have the production manager discover the night shift's stoppages the following day, on paper. The equipment is modern. The information chain is from the 90s. The real OEE stays hidden — and with it, the ability to react.

The recurring mistake: the company invests in visualisation platforms before resolving the integration between the shop floor and the ERP. They end up with two data silos, a pretty dashboard nobody consults, and an IT manager managing the discrepancy between systems instead of managing production. After 18 months, the platform is abandoned. The paper is back.

What you need before starting

Before assessing any technology, confirm you have these baseline conditions. Without them, any implementation is delayed or fails — and no supplier will tell you that in the commercial proposal.

You need a physical mapping of the lines: knowing which workstations, equipment and shifts you want to monitor. You need a production manager with real authority to change work orders in the system — not someone who has to ask the IT director for approval for every exception. You need an ERP or management system with an open API or a documented integration file; if your ERP supplier can't show you that within 48 hours, the problem runs deeper than it seems. You need an industrial network separate from the office network, or a concrete plan to create one before go-live.

There are two requirements companies systematically underestimate. The first: an agreed definition of "stoppage" — what counts as a breakdown, a micro-stoppage and a changeover. Without it, operators record what seems right to them and the data is not comparable between shifts. The second: management's commitment to act on the data that emerges, including the uncomfortable data. We see projects stall not through technical failure, but because the first real OEE report showed numbers management was not prepared to see — and nobody wanted to own the problem.

Step 1 — Measure the real OEE before buying anything

Calculate the current OEE manually over two weeks. Availability × Performance × Quality. Do it workstation by workstation, not by aggregated line — the line average hides the bottleneck.

What you'll find is almost always surprising. The first manual measurement reveals deviations of 15 to 25 percentage points from the internal perception. And there is a recurring pattern in the Portuguese factories we've worked with: the workstation flagged by management as "the problem" is often the one absorbing the inefficiencies of two upstream workstations. Attacking the visible symptom without measuring the complete system is the most expensive way of solving nothing.

Record OEE per workstation and per shift — not just per line. Define and agree the stoppage categories with the operators before you start measuring: an "other" category above 20% means the taxonomy is wrong. Record rejects per product reference, not just the total. Compare the planned cycle time with the actual cycle time: when the deviation is consistently above 12%, the problem isn't the operator — it's the planned cycle time that was never updated.

Step 2 — Identify where the information dies

Follow the flow of a production order from start to finish. Mark every point where information moves from digital to paper, or from paper to digital. Each transition is a point of loss — of time, of accuracy, of traceability.

In a typical garment factory in the Vale do Ave with subcontracting for international brands, it's common to have four or five transitions of this kind between cutting and dispatch alone. When the customer requests batch traceability for compliance with the EU Strategy for Sustainable and Circular Textiles, the answer takes days — because the information is fragmented across shift record sheets, Excel files from different supervisors, and the central ERP that only sees the finished product. The brand's auditor arrives, requests the history of a specific batch, and the quality manager spends three hours manually reconstructing what should be one click away.

The detail the manuals don't mention: the most critical point of loss is rarely the most obvious one. It's not the sheet of paper the operator fills in at the end of the shift — it's the Excel file the supervisor consolidates on Fridays and that never enters the ERP because "the ERP doesn't accept that format". That file is the real production management system. The ERP is the invoicing system with production fields left blank.

• All digital↔paper transition points in the production flow listed.
• The average latency between the real event and the record in the system identified.
• The data that exists only in the heads of key operators mapped — and the risk of those people leaving.

Step 3 — Prioritise by impact, not by technical ease

Use the following table to prioritise where to invest first. Assess each area by its impact on OEE and its implementation effort. Attack the high impact / low effort quadrant first — and resist the temptation to start with predictive maintenance just because it's the one that shows up at Industry 4.0 conferences.

Step 4 — Validate the integration before connecting the first sensor

The most expensive mistake we see repeated: the company installs industrial terminals, the data reaches the hardware supplier's own platform — and stays there. The ERP doesn't know what happened. The manager has two systems that don't communicate and ends up managing the discrepancy instead of managing production.

Before any purchase of hardware or a capture platform, answer these four questions. If you can't answer them all, the integration isn't specified — and you shouldn't proceed to purchase.

1. Will the captured production data automatically close the manufacturing orders in the ERP?
2. Will the stoppages recorded on the shop floor feed the OEE calculation without manual intervention?
3. Is the planned takt time accessible to the capture system for real-time comparison?
4. Who is responsible for resolving discrepancies between the capture system and the ERP — and what is the documented process?

A detail contracts rarely include: require that the integration specification be validated in writing by both suppliers — the ERP one and the capture platform one — before signing. When something fails in the integration, each supplier points at the other. With the specification signed by both, the perimeter of responsibility is defined. Without it, you're alone in the middle.

• Integration specification written and validated by both suppliers.
• Integration test carried out in a staging environment before go-live.
• Discrepancy reconciliation process documented and with a named owner.

Step 5 — Define the KPIs you'll manage, not the ones you'll report

There's a difference between management KPIs and reporting KPIs. The former change behaviour. The latter decorate board presentations.

For a medium-sized Portuguese industrial factory, five management KPIs are enough to move the needle. The OEE per critical workstation — not the factory average, the workstation that limits the flow. The reject rate per reference, with cause detail: the aggregate number is useless, because a rate of 3.2% hides the fact that a specific reference is at 11% and is consuming two shifts' worth of rework. The deviation of actual vs. planned cycle time per shift and per operator — when the deviation is consistent in one shift and not another, the problem is with the process, not the person. The mean time between failures (MTBF) on the three most critical pieces of equipment. And the on-time production order fulfilment rate, measured on the dispatch date — not on the manufacturing completion date, which is how most factories measure it and which systematically conceals the time lost between production and dispatch.

Set up automatic alerts for these five with action thresholds, not just information thresholds. And make them visible on the shop floor — on a screen in the line's aisle, on a workstation terminal. A KPI that only exists in the management office doesn't change the behaviour of those who produce.

Common mistakes — and what really causes them

Buying an IoT platform without specifying the integration with the ERP. The result is two data silos, manual reconciliation and abandonment of the platform within 18 months. The real cause isn't technical incompetence — it's that the purchasing process was led by the IT department based on product demos, without the production manager in the room and without the ERP supplier in the conversation. Fix it: the integration specification is a condition of award, not an annex for later.

Launching the Industry 4.0 project without reasonably reliable historical production data. The anomaly detection algorithms learn the wrong patterns and generate false alerts. After three weeks, operators ignore all alerts — including the genuine ones. Do two months of structured manual capture before connecting any automatic system.

Ignoring the operators in the design of the capture process. The terminals end up being used to record what "looks good" in the system, not what happened. Involve two or three key operators in the pilot. What they say doesn't work is exactly what will fail in production — and they know it before any consultant does.

Measuring aggregated OEE instead of OEE per workstation. A factory with an average OEE of 72% may have a critical workstation at 48% that's limiting the whole line. The average not only hides the bottleneck — it gives management a false sense that things are reasonable.

Treating Industry 4.0 as an IT project instead of an operational project. The IT manager delivers the system; the production manager doesn't adopt it. The project sponsor has to be the operations director. IT is an implementation partner, not the owner of the outcome. When the sponsor is IT, the project is assessed by uptime and closed tickets. When the sponsor is operations, it's assessed by OEE and deadline compliance. These are completely different metrics — and they produce completely different projects.

Where preparation ends and implementation begins

If you've reached this point with the five KPIs defined, the integration specified by both suppliers and OEE per workstation calculated for the last two weeks, you're at the level of preparation that allows an implementation without fundamental surprises. Most Portuguese factories aren't there yet — and not for lack of will, but because nobody told them this was the work to do before buying the first terminal.

The topic of production capture and OEE calculation is covered in more detail in the operational guide MES and OEE in the Portuguese factory. If your factory operates in a network with subsidiaries or subcontractors, the issue of integration between shop-floor capture and central management — without data duplication and without manual reconciliation — is dealt with directly in KORA Productivity together with the MULTI ERP. To understand what to ask your ERP supplier before proceeding, read the guide on industrial ERP vertical fit.

Sources

• CEFAMOL — Associação Nacional da Indústria de Moldes. Indústria Portuguesa de Moldes: Dados de Setor 2023. Marinha Grande, 2023. Available at: www.cefamol.pt
• AIMMAP — Associação dos Industriais Metalúrgicos, Metalomecânicos e Afins de Portugal. Metal Portugal: Dados do Setor. Lisbon, 2023. Available at: www.aimmap.pt