MES and OEE in the Portuguese factory: a complete operational guide

At a garment factory near Famalicão, the production manager knows how much was produced yesterday. Ask him how much he lost — in stoppages, micro-stoppages, rework, thread changes — and the answer is a shrug. "It went well." That "it went well" costs him between 25% and 40% of capacity that he pays for in rent, energy and wages but never sees. The problem is not that the factory works badly. It is that nobody measures where the time disappears.

This guide deals with the one metric that makes that loss visible and actionable — OEE — and with the software layer that captures it without lying: the MES. The thesis is simple and uncomfortable: without automatic capture on the shop floor, any efficiency figure your company reports is accounting fiction.

We will work through the operational problem, what MES and OEE mean exactly, the Portuguese landscape, the implementation models, how to diagnose your readiness, the regulatory framework you inherit the moment you connect the first machine to the network, and a quarterly roadmap that delivers a defensible figure. No slide-deck theory. With the factory we know.

1. The real operational problem

Most Portuguese industrial SMEs measure production. Almost none measure efficiency. The difference is not semantic.

Production is what goes out the door: pairs of shoes, metres of knit, injected parts. Efficiency is the relationship between what went out and what could have gone out with the equipment, time and labour you paid for. A sewing machine can be "switched on" for eight hours and produce the equivalent of four. The other four evaporate in thread that snapped, in the operator who went to fetch components from the warehouse, in the model change that took 50 minutes instead of 20, and in the parts that went back for rework because the seam came out crooked.

When we ask an industrial manager "what was last week's OEE", the honest answer is almost always one of two: "we don't measure it" or "I've got a sheet here". The sheet, when it exists, was filled in at the end of the shift, from memory, with figures rounded down on the losses and up on the production. It is a morale report, not an operational one.

The invisible cost by sector

Each Portuguese industrial vertical hides the loss in a different place. Knowing where to look is half the job.

• Textiles (Vale do Ave) — stoppages in spinning and weaving due to thread breakage, article changes, and setting-up time on finishing machines. A circular knitting machine idle for 30 minutes per shift due to unrecorded micro-stoppages loses more than 200 hours a year that nobody accounts for. In a hall with 20 looms, that pattern multiplies until it becomes a phantom production line that never appears in the budget.
• Footwear (Felgueiras, Guimarães, S. João da Madeira) — SKU complexity kills efficiency in series changes. A collection with 1000 references and three axes (colour, size, last) means the line is constantly changing. Every minute of poorly planned setup multiplies by the hundreds. The pressure of the two annual visits from international buyers — men's footwear in August, women's in February — compresses everything into windows where efficiency decides whether the order ships on time.
• Garment/apparel manufacturing — the bottleneck is line balancing. Unbalanced operations create intermediate stock and waiting times that appear nowhere until the parent company (Inditex, Decathlon, Lacoste) complains about the lead time. With an ageing workforce and long learning curves on new models, the imbalance is structural and rarely measured.
• Metal/plastics (Aveiro–Marinha Grande corridor) — injection moulds with theoretical cycles of 32 seconds that run at 41 in practice. The 9-second difference per cycle, in series production, is money trickling away. On a machine running 1500 cycles per shift, that is more than three and a half hours of paid machine time producing below nominal — every day.
• Distribution and retail — the concept of OEE migrates from the equipment to the picking station and the checkout line. The warehouse manager in the Lousada/Paços corridor who loses 40 minutes per shift on poorly routed searches has the operational equivalent of a machine running slow. The capture logic is the same; the sensor changes.

If you cannot say to the cent how much your last series change cost you, you don't have a production problem — you have a measurement problem.

Why Excel is no longer enough

The average Portuguese industrial manager runs the factory with a spreadsheet that someone fills in at the end of the shift, from memory, with the figures rounded to what "seems right". That file has three fatal flaws: it is retrospective (you find out about the problem the next day), it is manual (the operator doesn't record the 4-minute micro-stoppage that happened 12 times), and it is negotiable (the supervisor is not going to report that his line ran at 60%).

There is a fourth flaw, less obvious: Excel doesn't scale. It works for one line and one shift. For five lines, three shifts and two factories, it turns into an archipelago of files with diverging versions that nobody consolidates without losing a morning. And when the international client asks for the traceability of a batch, Excel doesn't have the granularity — it knows how much was produced in the day, not on which machine, with which roll of thread, at what exact hour.

The MES solves all four. Capture in real time, automatically, with no room for diplomacy, and with the granularity that the audit demands.

What "it went well" hides — the three types of loss

The supervisor's intuition is not stupid — it is incomplete. He feels the big breakdowns, because they stop the line visibly. What he doesn't feel are the two losses that hurt most in the aggregate:

• Micro-stoppages. Four minutes to unjam, two to restock, three to readjust. Individually, noise. Added up over a shift, they can eat 15% to 20% of availability. No operator notes them down — it's not worth picking up the notebook for four minutes.
• Running slow. The machine works, but at 80% of nominal speed because the material is damp, the tool is worn or it simply "always ran like that". It is the most invisible loss of all, because the machine appears to be functioning.
• Scrap at source. Parts that come off the machine already out of specification and are only caught at final inspection, after having consumed machine time, material and labour. The later it is caught, the more expensive it is.

These three losses are exactly the three denominators of OEE. It is no coincidence: the metric was designed to illuminate precisely what intuition doesn't see.

2. What exactly are MES and OEE on the factory floor

Two concepts that go together but are not the same thing. Confusing them is the number one mistake made by those who buy industrial software.

OEE: the metric

OEE stands for Overall Equipment Effectiveness. It is a figure between 0% and 100% that results from multiplying three factors:

The formula: OEE = Availability × Performance × Quality. Each factor is a percentage. If availability is 85%, performance 90% and quality 95%, the OEE is 0.85 × 0.90 × 0.95 = 72.7%.

The danger lies in the individual figures all looking good. 85%, 90%, 95% — any manager would sign off on that. Multiplied, they give 73%, which means more than a quarter of capacity is being wasted. That is why OEE deceives: the loss hides in the multiplication. Consult the full definition in our glossary of industrial terms for the adjacent concepts.

The six big losses — the classic anatomy

The TPM tradition breaks OEE down into six big losses, distributed across the three factors. Knowing them gives you the taxonomy of stoppage reasons before configuring any terminal:

Each of these losses has a different operational owner. Breakdowns are maintenance. Setup is planning and method. Micro-stoppages are internal logistics and station ergonomics. Running slow is preventive maintenance and material quality. When OEE drops, the breakdown tells you whose door to knock on — it doesn't force a witch-hunt.

An OEE of 85% is considered world-class in discrete industry. The average Portuguese factory that has never measured typically runs between 45% and 65% — and has no idea.

MES: the system that measures

MES stands for Manufacturing Execution System. It is the software layer between the ERP (which knows what should be produced) and the shop floor (where production happens). The MES captures what really goes on: which order is in progress, on which machine, at what speed, with what stoppages, with what scrap.

Without an MES, OEE is calculated from data that someone wrote by hand. With an MES, OEE calculates itself from what the sensors and terminals recorded. It is the difference between the blood pressure the patient says he has and the one the device measures.

A competent MES does more than count parts and time stoppages. It manages the sequence of production orders, records the actual consumption of material against the theoretical, keeps the batch trail (which roll, which dye, which operator, which shift), and returns the reality of production to the ERP to close costs. It is the nervous system of the factory — it captures the stimulus at the station and carries it up to the management decision.

MES, MRP, ERP, SCADA: the map

The confusion of acronyms sinks projects. Here is the hierarchy:

• ERP — plans and records the business: orders, purchasing, invoicing, finance. The ERP MULTI verticalises this for industry.
• MRP — within the ERP, calculates material and capacity requirements. It says what and when to produce.
• MES — executes and monitors production in real time. It says how it is going now.
• SCADA/PLC — controls the machine at the electrical level. The MES reads from it, it doesn't replace it.

The MES feeds off the ERP (receives the production orders) and returns reality to it (actual production, consumption, times). It is this bidirectional link that separates a serious capture system from a glorified parts counter. The layer INFOS dedicates to this is KORA Productivity.

The standard that arranges these levels is called ISA-95. It defines five functional levels, from the physical control of the equipment (level 0-1) to business planning (level 4), with the MES occupying level 3 — the border where execution talks to management. It matters because any serious integrator designs the architecture with this layer in mind; those who ignore the levels end up connecting the ERP directly to the PLC and praying the network holds.

A brief history — why this matters now

The concept of OEE was born in Japan in the 1960s-70s, within Toyota's and Nippon Denso's TPM (Total Productive Maintenance). It reached the West in the 1980s. The MES as a software category was formalised in the 1990s with the MESA-11 model and, later, with the ISA-95 standard, which defines how the execution level converses with the management level. In Portugal, real adoption in SMEs only accelerated in the last decade — pushed by two factors: international clients demanding traceability, and EU funds (PT2030, PRR) financing the digitalisation of the shop floor.

There is a third force, quieter: the shortage of labour. As it becomes harder to hire experienced operators and supervisors, the knowledge that lived in the head of the "man who knows the machine" can no longer be the only source of truth. The MES codifies that knowledge — cycle times, stoppage reasons, setup sequences — so that it doesn't walk out the door on retirement day. In garment factories with an ageing workforce, this has ceased to be optional.

3. The landscape in Portugal today

Manufacturing weighs significantly on the Portuguese economy, but its digitalisation is uneven. The large exporting factories already have MES. The SMEs — the overwhelming majority of the national industrial fabric — still run on spreadsheet and instinct.

Where we stand in the European index

The European Commission's Digital Economy and Society Index (DESI) consistently places Portugal below the EU average in the integration of digital technologies in companies. Portuguese SMEs adopt basic management tools faster than advanced production technologies — ERP and electronic invoicing came in (partly by legal obligation), but MES, industrial IoT and production analytics remain a minority.

The honest reading is that Portugal first digitalised the part the State mandated (SAF-T, ATCUD, certified invoicing) and the part of least friction (email, web presence, basic ERP). The digitalisation that requires capital investment, process change and shop-floor management — exactly where the MES lives — was left behind. It is terrain where the national fabric has enormous room to catch up, and it is precisely there that the digital transition funds point.

Qualitative reading based on DESI patterns and market reality; these are not official percentages by tier.

The portrait by company size

The difference is not just budget — it is decision structure and process maturity.

The small company (10-50 employees) decides fast, usually at the owner. It has little formalised process and the "how it's done" lives in the people. The MES here is a double-edged sword: the gain is enormous (it goes from zero measurement to an objective figure), but change management is intimate — everyone knows everyone, and measuring is personal. Starting small, with one station, is mandatory.

The medium company (50-250) already has ERP, usually has a dedicated production manager and, often, that self-taught IT professional with 15 years of business knowledge and no formal qualification — the hero who makes everything work. It is the profile where the MES gives the best return, because there are identifiable bottlenecks, there is volume to justify the investment, and there is someone who understands the data. The risk is the ego of "we already know where we lose" stalling the measurement.

The large company and the multi-factory group already have MES somewhere, but frequently in islands — each unit with its own taxonomy, its own sheet, its own system. The challenge stops being to capture and becomes to compare. Without harmonised stoppage reasons across factories, one factory's OEE doesn't speak to the other's, and the consolidation in Qlik Sense shows apples and oranges.

The pressure from international clients

Portuguese textiles and footwear live off subcontracting for European parent companies. Inditex, Decathlon, Mango, Lacoste, Tom Tailor — all have raised their traceability requirements. The EU Strategy for Sustainable and Circular Textiles and the future Digital Product Passport will make batch traceability mandatory: which thread, which dye, which origin, which process. Those who cannot prove the chain don't stay on the vendor list. We discuss this in detail in the context of the textile sector.

The smell of the Vale do Ave dye houses is not nostalgia — it is compliance. When the brand asks for proof that that batch of knit was dyed with a given process, on a given date, with a given dye supplier, the factory that has that information scattered across notebooks and in the head of the dye-house supervisor loses the order to the one that has it in a system. Automatic traceability has ceased to be a differentiator. It has become a ticket to entry.

Traceability has ceased to be a compliance luxury. It is the condition of market access. And traceability without automatic production capture is manual work doomed to fail the audit.

The financing that changes the equation

The PRR and the PT2030 (with COMPETE 2030 and Norte 2030) finance industrial digitalisation. MES, IoT and Industry 4.0 projects fall within digital transition calls. In practice, what we see is that projects with a clear operational use case and verifiable metrics (OEE before/after, setup reduction, scrap reduction) pass; those that sell abstract "digital transformation" get stuck in the technical reports. Structure the application around a figure you can measure.

The recurring mistake in the application is confusing the assessor with the engineer. The technical assessor does not want to read that you are going to "leverage Industry 4.0 synergies". He wants to read that line X's OEE is at 54%, that the objective is 68% in 12 months, that this frees up the equivalent of Y machine hours per year, and that the investment pays for itself in Z months. Concrete figures, measured baseline, defensible objective. The applications we handle with the client always start from the manual measurement he did beforehand — because without a baseline there is no demonstrable gain, and without a demonstrable gain the report gets stuck.

The calendar nobody controls

There is a seasonal rhythm the software has to respect. In footwear, the collections and buyer visits dictate production peaks in which nobody accepts a rollout. In textiles and garment manufacturing, the "month-end" and the parent-company deadlines compress everything. The warehouse manager in the Lousada/Paços corridor will fight any implementation that takes him off the radar for more than two hours at a stretch — and he is right, because during those two hours the warehouse doesn't stop receiving and dispatching. Planning capture around these rhythms, and not against them, is the difference between adoption and sabotage.

4. The implementation models

There are four approaches to getting MES and OEE working in a Portuguese factory. They are not equally good — they depend on your size, on the heterogeneity of the machine park, and on the maturity of the team.

Model 1: Manual capture assisted by terminal

Operators record stoppages, reasons and quantities on industrial terminals (rugged tablets, touch screens at the station). There are no sensors — the person presses the button "stopped due to breakdown", "stopped due to series change", "stopped due to lack of material". It is the cheapest point of entry and the fastest to get going.

Good for: factories with many manual stations (garment manufacturing, sewing, footwear assembly) where the machine cannot be read automatically. Risk: it depends on recording discipline. Without management, operators always press the same button. The quality of the data is only as good as the habit created — and the habit is created with the team leader using the figure in the morning meeting, not with a memo.

Model 2: Automatic capture through sensorisation (IoT)

Sensors and connections to the PLC read the machine directly: cycle counting, state (producing/stopped), speed. The operator only intervenes to classify the reason for the stoppages. It eliminates the subjectivity of counting.

Good for: homogeneous and automated machine parks (weaving, plastic injection, spinning). Risk: integrating old machines with proprietary PLCs or without any digital output. Marinha Grande is full of 20-year-old injection machines that don't talk to anyone. For these, there is always the option of non-intrusive external sensors (cycle counters, vibration or electrical consumption detection) — cheaper than replacing the machine, and sufficient for OEE.

Model 3: Hybrid (sensor + terminal)

The sensor counts and detects the stoppage; the operator classifies it on the terminal. It is the model we recommend for most Portuguese industrial SMEs because it combines the objectivity of the sensor with the context that only the person gives. The sensor says "stopped at 14:32 for 6 minutes"; the operator says "it was a series change". The machine doesn't lie about the how much; the person explains the why.

Good for: practically all mixed parks. Risk: overloading the operator with classifications. Keep the taxonomy short (8-12 reasons) and the buttons large — whoever is classifying has dirty hands and is in a hurry.

Model 4: MES integrated into the ERP vs. standalone MES

The structural decision: does the MES live inside the ERP ecosystem or is it a separate product that integrates afterwards?

For Portuguese SMEs, native integration wins almost every time. When the MES and the ERP are from the same ecosystem — like KORA Productivity linked to the ERP MULTI — the production order, the material consumption and the recorded production live in the same data model. There is nobody pointing the finger at the integration when the numbers don't match. For operations of the highest complexity, QAD Adaptive ERP offers the depth of execution that multinationals demand.

The hidden cost of standalone integration

Those who choose the separate MES rarely account for the cost of maintaining the bridge. Every ERP update risks breaking the interface. Every MES update likewise. When the production figures don't match those of the stock, the forensic investigation begins: did the MES count wrong, did the synchronisation fail, did the operator classify incorrectly? In factories with a one-person IT department — that self-taught hero — this maintenance load falls entirely on him, and it is time he doesn't have. Native integration is not only cheaper; it is one less thing that can break at three in the morning on month-end close.

5. How to assess whether your company needs it

Not every factory needs an MES tomorrow. But every factory that doesn't know its OEE is losing money it can't see. Use this diagnostic.

Signs that you need it, and now

• You cannot say, without investigating, what last week's OEE was by line.
• Capacity decisions ("do we accept this order?") are made by the supervisor's instinct.
• An international client has already asked you for batch traceability and you had to put it together by hand.
• Series changes take times that nobody clocks but everyone suspects are high.
• Scrap and rework appear at the end of the month as an unpleasant surprise.
• You have an application, or intention to apply, for digital transition funds without a measurable use case.
• When there is a peak of orders, nobody knows whether the factory has real slack or is already at the limit.
• The knowledge of the times and tricks of each machine is in the head of one or two people close to retirement.

Signs that you can wait

Honesty requires it: there are factories where the MES is not the number one priority. If you don't yet have a serious ERP, or what you have is a generalist product that doesn't model your colour-size-last axes, shop-floor capture is orphaned — there is nowhere to land the production order. Solve the foundation first. Assess the vertical fit of your industrial ERP before instrumenting stations. Capturing production in real time to dump it into an ERP that doesn't know how to use it is spending money on telemetry that nobody reads.

Step by step: MES/OEE readiness diagnostic

1. Map the bottlenecks. Identify the 3-5 machines or lines that limit the factory's capacity. Don't measure everything — measure where it hurts. Applying an ABC Analysis to the work centres tells you where the capture effort pays off best.
2. Time a series change, by hand, three times. With a stopwatch, now. The value will shock you and give you the "before" baseline that justifies the project.
3. Define the stoppage reason. List 8 to 12 stoppage codes that make sense for your factory (breakdown, change, lack of material, lack of operator, adjustment, meal break, etc.). Without a taxonomy of reasons, OEE doesn't tell you where to act.
4. Assess the machines' readability. For each critical piece of equipment, check whether it has a digital output (accessible PLC, counter) or whether it will have to be manual capture. This defines the implementation model.
5. Calculate the OEE of one line, manually, for a week. Yes, by hand, with a notebook at the station. The figure that comes out is your current truth and the argument that convinces the CFO.
6. Quantify the cost of the loss. Multiply the lost hours by the real hourly cost of the station (operator + machine + overhead). That figure is the budget the project can justify.
7. Assess the team's maturity. Does the team leader read figures or just feel the factory? Is there anyone who uses the data in the morning? Without this person, the best MES turns into a dead archive.

Before buying any OEE software, measure one line by hand for a week. Whoever can't manually measure one line won't know how to configure the capture of a hundred.

Quick wins before the project

Three things you can do in one to two weeks, without investing in an MES:

• Post at the station the target time for each series change and clock against it. Just the visible measurement already reduces the time.
• Standardise the taxonomy of stoppage reasons across the whole factory — even if recorded on paper — so the data is comparable when the MES arrives.
• Choose one pilot line and calculate its weekly OEE manually. It will create the internal appetite for the figure.

These three steps cost no software and do two things: they give you the baseline you will need for the fund application and for the CFO, and they test whether your organisation has the discipline to use data. If the manual sheet lasts two weeks and then dies, the problem is not technological — it is a management one, and no terminal will solve it.

6. What to choose and why (decision by company size)

The right question is not "which is the best MES" — it is "which is the best MES for a factory like mine". Here is the matrix.

Start with the bottleneck, not the whole factory

The most expensive mistake we see is the factory that decides to instrument 100 stations at once. The project becomes an 18-month monster, the team burns out, and the ROI disappears over the horizon. Start with 3-5 machines that limit capacity. Prove the figure. Extend afterwards.

There is a logic behind this that goes beyond prudence. The bottleneck, by definition, dictates the pace of the entire factory. An hour gained at the bottleneck is an hour gained in total production; an hour gained on a machine that is not a bottleneck adds nothing, because the part will wait in the bottleneck's queue anyway. Measuring and improving the bottleneck first is where every euro of capture yields most. The other machines come next, in order of impact on capacity.

OEE doesn't live alone — it needs BI

Capturing the OEE is half. The other half is looking at it in a way that lets you decide. The MES data feeds dashboards where the manager sees trends, compares shifts and lines, and identifies stoppage patterns. It is here that Qlik Sense turns capture into decision, with Self-Service BI that the industrial manager uses without depending on IT for every chart.

The difference that BI brings is the move from the figure to the pattern. An OEE of 62% on one day says little. The OEE trend over three months, segmented by shift, shows that the night shift runs systematically 8 points below — and there you already have a concrete conversation to have. Cross-referencing stoppage reasons with the time of day reveals that micro-stoppages spike after lunch. It is these patterns, and not the isolated figures, that change decisions.

An MES without an analysis layer is a black-box recorder that nobody listens to. The value is not in capturing the OEE — it is in someone looking at it every morning and changing something.

The human factor: the tape on the tablet

A warning from someone who has already implemented: the operator who doesn't want to be measured will sabotage the system. We have seen terminals "strategically" covered, sensors "that broke down on their own", and the classic KORA tablet hidden behind a box with tape. Production capture is as much a change-management project as a software project. If the team leader doesn't buy into the idea, no hardware will save it. Involve him from the diagnostic — and show him that the figure protects him as much as it exposes the line.

The argument that works with the team leader is not "we're going to monitor your line". It is "when the manager asks you why the line ran badly, you'll have the figure that shows it was the machine broken down for three hours, not your people marking time." OEE well used defends those who work well and exposes the structural problems that are nobody's fault on the floor. When the supervisor understands this, he stops covering the terminal and starts using it to justify himself — which is exactly what you want.

The risk of measuring badly and using it to punish

The flip side of the coin: a company that uses OEE to chase operators destroys the reliability of the data in two weeks. People learn to classify stoppages in the way that protects them, not in the way that is true. The figure becomes theatre. The rule we defend is simple — OEE measures the system, not the person. Attack the process, the method, the maintenance, the logistics. The moment the figure enters individual appraisal carelessly, you lose the data. And, as we will see, there are legal limits to that use.

7. Applicable regulatory framework and compliance

MES and OEE are not, in themselves, regulated matter. But the system that captures them touches data, networks and processes that are under various Portuguese and European legal obligations.

Traceability and the EU Textile Strategy

The EU Strategy for Sustainable and Circular Textiles and the Ecodesign for Sustainable Products Regulation (ESPR, Regulation EU 2024/1781) introduce the Digital Product Passport. For Portuguese textiles and footwear, this means mandatory batch traceability in the near horizon. The MES is the infrastructure that makes that traceability automatic instead of a manual audit exercise.

The Digital Product Passport will be phased by product category, and textiles are among the priority ones. Those who capture production with a batch trail from now build the database the passport will require, instead of having to reconstruct it in a rush when the obligation kicks in. It is a rare case where anticipating the regulation coincides exactly with gaining efficiency — the same capture serves both ends.

Cybersecurity: NIS2 and the connected factory

As soon as you connect machines to the network for production capture, the factory becomes an attack surface. Business Continuity ceases to be theoretical when a ransomware encryption can stop production. The NIS2 Directive (EU 2022/2555), transposed into Portugal by DL 65/2025, extends cybersecurity obligations to many medium-sized industrial companies. We deal with this in detail in the guide to NIS2 cybersecurity for Portuguese industrial SMEs. The separation between the production network (OT) and the office network (IT), supported by architectures such as SD-WAN, is the first line of defence.

The point many factories underestimate is that the production network has different requirements from the office network. The capture terminals and sensors live in an OT (Operational Technology) environment, which was traditionally designed for isolation, not for internet connection. Connecting that network without proper segmentation is opening a direct door from the web to the heart of the factory. ENISA, the European cybersecurity agency, has been pointing to IT/OT convergence as one of the fastest-growing risk vectors in industry. Design the segmentation in the capture project, not as a later patch.

GDPR and operator data

The MES records who produced what, when, and at what pace. That is personal performance data. The GDPR and Law 58/2019 apply: define the purpose, minimise the collection, inform the workers, and take redoubled care if you use the figure for individual appraisal — sensitive territory in labour relations. Individual performance monitoring has legal limits that your DPO must validate.

In practice, there is a boundary that Portuguese labour law takes seriously: the difference between measuring the process and surveilling the person. Capturing a line's OEE is production management. Using that OEE to build an individual ranking of operators and take disciplinary decisions is performance monitoring, with obligations of information, proportionality and, in many cases, consultation. Define the purpose on paper, communicate it to the workers, and keep the collection proportional to the declared objective. In companies with more than 50 employees, also coordinate with the whistleblowing channel mandated by Law 93/2021, should the data processing be contested internally.

Predictive AI and the AI Act

When the next step is predictive analytics — anticipating breakdowns, predicting scrap — the AI Act (Regulation EU 2024/1689) comes into play. Most predictive maintenance use cases are minimal risk, but any application that touches the appraisal of people rises on the risk scale. Classify before implementing.

The simple mental rule: AI that predicts the behaviour of a machine is, in general, minimal or limited risk. AI that appraises, classifies or decides about people (individual productivity, shift allocation based on predicted performance) rises on the scale and drags in obligations of transparency, human oversight and documentation. Predicting a mould's breakdown is one thing; the algorithm that decides who stays and who goes based on predicted productivity is another, and it requires serious conformity assessment before touching the production button.

ISO 27001 and the integrity of production data

If your factory certifies processes or serves clients who demand information security guarantees, ISO 27001 structures the governance of production data — access control, integrity, audit trails. The OEE figure you report to the client has to be as reliable as the invoice.

The complementary standards — ISO 27017 for cloud services and ISO 27018 for personal data in the cloud — gain relevance when production capture lives partially or wholly in hosted infrastructure. The question the international client will ask in the supplier audit is not only "what is your OEE", but "how do you guarantee that this figure was not tampered with and that my batch data is secure". Having the certifications structured turns that conversation from an obstacle into a commercial argument.

The moment you connect the first machine to the network to capture production, you have inherited a cybersecurity problem. Deal with it in the design of the project, not after the first incident.

Invoicing, SAF-T and coherence with the shop floor

The production recorded in the MES feeds the costing that, in the end, underpins the invoicing. DL 28/2019 and Ordinance 195/2020 impose certified electronic invoicing, ATCUD and monthly SAF-T reporting. When the captured material consumption and production times don't match what the ERP invoices, an inconsistency opens up that both accounting and the AT can question. The native integration between capture and ERP closes this circle: what was produced, what was consumed and what was invoiced tell the same story.

8. How INFOS approaches this

We have built vertical software for industry for more than three decades, and shop-floor capture is one of the pieces that most distinguishes an SME that knows where it loses money from one that guesses. KORA Productivity captures production in real time, calculates OEE by machine, line and shift, and classifies stoppages with the taxonomy your factory defines — on industrial terminals designed to survive dust, oil and the rush of the end of the shift.

The difference we defend is native integration. KORA Productivity links to the ERP MULTI, which means the production order, the material consumption and the recorded production share the same data model. There is no fragile interface to synchronise, nor the "it's the integration's fault" game when the numbers don't match. And because we know Portuguese footwear, textiles and the metal-plastics industry, the modelling of SKUs, colour-size-last axes and production routings is not a painful adaptation of a generic product — it is what the system already does.

What we have learned to do — and not to do

Thirty-five years of projects teach through failure as much as through success. The capture implementations that went wrong nearly always had the same signature: they started too big, without a manual baseline, and without the team leader involved. The terminals stayed there, pretty, but the data was never used in the morning, and within three months the project was a dead archive that everyone pretended didn't exist.

The ones that went well were small, measured by hand before automating, and had a manager who took up the figure and asked questions with it. That is why our approach is not to sell hardware — it is to install measurement discipline at a station that matters and to prove the figure, before any extension. The technology is the easy part. Change management is where projects are won or lost.

Frequently asked questions

What is OEE and why is it critical to measure in Portugal?

OEE (Overall Equipment Effectiveness) measures the real efficiency of the equipment: actual production versus theoretical possible production. In Portugal, most SMEs report production, not efficiency. The difference costs 25% to 40% of capacity paid for but invisible — stoppages, micro-stoppages, rework and series changes that nobody systematically accounts for.

What is the difference between production and efficiency?

Production is what goes out the door: pairs of shoes, metres of knit, parts. Efficiency is the relationship between what went out and what could have gone out with the same equipment, time and labour. A machine can be switched on for eight hours and produce the equivalent of four — the other four are lost in breakages, model changes and rework.

Why does Excel no longer work for measuring efficiency?

Excel is retrospective (you find out about the problem the next day), manual (micro-stoppages are not recorded) and negotiable (supervisors don't report the truth). It doesn't scale for multiple lines and shifts, it creates diverging files and it doesn't have the granularity for batch traceability that international clients demand.

What is the invisible cost of inefficiency in Portuguese textiles?

In Vale do Ave, the losses concentrate in thread breakages, article changes and machine adjustment. A circular knitting machine idle for 30 minutes per shift loses more than 200 unaccounted hours a year. In a hall with 20 looms, this is equivalent to a phantom production line that never appears in the budget.

How does Portuguese footwear lose efficiency?

SKU complexity kills efficiency in series changes. Collections with 1000 references mean constant changes. Every minute of poorly planned setup multiplies by the hundreds. The pressure of the annual visits from international buyers compresses everything into windows where efficiency decides whether the order ships on time.

What is an MES and how does it solve the problem?

MES (Manufacturing Execution System) is the software layer that automatically captures data on the shop floor without lying. It solves the four flaws of Excel: capture in real time, automatically, with no room for diplomacy, and with granularity for audit. It makes visible where the time disappears.

What are the three types of loss the supervisor can't feel?

Micro-stoppages (four minutes unjamming, two restocking, three readjusting) add up to 15% to 20% of daily availability. Running slow (machine at 80% of nominal speed due to damp material or worn tool) costs accumulated hours. Rework due to deficient quality is invisible until the client complains about the lead time.

Sources

• Standard ISO/IEC 22400-1:2014 — "Automation systems and integration — Key performance indicators (KPIs) for manufacturing operations management" — Defines operational efficiency metrics including OEE in the industrial environment
• Standard ISO/IEC 22400-2:2014 — "Automation systems and integration — Key performance indicators (KPIs) for manufacturing operations management — Part 2: Guidance