Business KPIs: the operational guide for Portuguese industrial directors

In a meeting room near Guimarães, an industrial director opens a 14-page report printed the previous Friday. He is looking at Monday's figures. Thursday's production has already changed everything. He knows it, the CFO knows it, and nobody says anything because that is how it has always been done. This is the problem with business KPIs in Portugal: it is not that companies don't have them — it is that they have them late, scattered, and misaligned with what the shop floor actually does.

The thesis of this guide is simple and uncomfortable: a KPI that arrives after the decision has to be made is not a performance indicator — it is an obituary. And most Portuguese industrial SMEs measure the past with a precision they do not apply to managing the present.

We will work through the problem from its root — because it is operational before it is technical — to a 90-day plan that any company in textiles, footwear, distribution or metal can execute without stopping production. Along the way, real Portuguese data, shop-floor examples you will recognise, and the decisions that separate a useful dashboard from a hard drive full of screens nobody opens.

1. The real operational problem

Let us be concrete, because the subject attracts generalities the way cotton attracts dust. The problem is not a lack of data. Portuguese factories generate data everywhere: production sheets, transport notes, work orders, clocking records, invoicing. The problem is that this data lives in silos that don't talk to each other and reaches management as blurry photographs of a train that has already gone.

The report nobody uses to decide

In a garment factory in the Vale do Ave with 80 employees, the typical end-of-month ritual involves someone — often the self-taught IT hero with 15 years at the company — cross-referencing Excel sheets for two days to produce a margin-by-order report. By the time that report is ready, the following month's orders have already come in, the prices from the parent company (an Inditex, a Decathlon, a Tom Tailor) have already been settled, and the report only serves to confirm what was already suspected: that reference is loss-making. But there is no longer any scope to renegotiate.

This is not incompetence. It is architecture. When indicators are rebuilt manually after the fact, the cost of producing them is so high that nobody produces them frequently enough to act. Calculation becomes an act of archaeology — digging up what happened — when it should be an act of navigation — deciding where we are going.

There is a hidden cost in this architecture that rarely enters the reckoning: the opportunity cost of the very person doing the cross-referencing. In an SME of 80 people, the «IT hero» is frequently the person who knows the business best end to end — from the purchasing circuit to the costing logic of a reference. Spending two days a month pasting cells is spending the company's scarcest resource on the most automatable task that exists. When that person finally leaves or falls ill, the margin report disappears with them, and the company discovers that it did not have a process — it had a person.

The KPI that measures the wrong thing

There is a more subtle and more expensive error. In a footwear factory in Felgueiras, the reigning indicator tends to be pairs produced per day. It is a number that goes up and pleases everyone. The problem: a sample collection has between 800 and 1,200 SKUs, spread across three axes (colour, size, last), and the value is not in the volume — it is in the ability to meet the date on which the international buyer visits the factory (men's footwear in August, women's in February). Measuring pairs/day in a sample environment is measuring the speed at which you make mistakes.

The same pattern repeats in every sector with a different face. In textiles, the misleading KPI is the number of metres produced per shift, when what the parent company penalises is the shade deviation between dyeing batches — a quality problem that volume hides. In distribution, it is lines dispatched per day, when what loses customers is the picking error rate: one wrong line in a delivery of building materials comes back as a return, a complaint and a furious call to the sales rep. Each of these indicators has one thing in common — it is easy to measure, it goes up consistently, and it measures activity rather than outcome.

An indicator that always goes up is almost always the wrong indicator. A good KPI bothers someone.

The test question is simple: if this number doubled tomorrow, would the company be better off? If the answer is «it depends», the indicator is measuring effort, not value. Pairs/day can double with a collection full of errors. Metres/shift can double with colour deviations that generate returns. Lines dispatched can double with hurried, incorrect picking. The KPIs that pass this test are few, and they are precisely the ones nobody wants to look at because they expose real problems.

The misalignment between who measures and who decides

Then there is the cold war between finance and operations. The CFO wants margin, DSO, stock turnover. The COO wants OEE, on-time delivery, defect rate. Both are right, and both measure realities that do not close on the same sheet. We explored this tension in detail in Industrial KPIs the CFO and COO disagree on — and how to align them, because it is where most BI projects derail: not through technical failure, but through a lack of agreement on what «good» is.

The classic case: the COO celebrates an OEE of 82% on a line, and the CFO points out that that line produced to stock a reference that turns over once a year. The machine was extremely efficient at manufacturing money sitting idle in the warehouse. Both indicators are correct. What is missing is the third number that reconciles them — efficiency weighted by real demand, or margin generated per machine-hour. Without that link, each area optimises its own KPI and the company as a whole loses.

Where the misalignment is born: the dual source of the data

The technical root of this conflict is almost always the same: finance and operations drink from different sources. Finance reads from the ERP's financial module; operations read from production sheets, isolated terminals, or nothing. When the cost of a work order that finance records does not match the hours the supervisor noted in his notebook, nobody knows which is right — and the meeting turns into a courtroom instead of a decision-making committee. The cure is not to side with one party. It is to ensure that both numbers descend from the same source of truth.

2. What exactly are «business KPIs» in Portugal

A KPI — Key Performance Indicator — is a quantified metric, associated with an objective, with a target and an owner. Every word of this definition eliminates candidates. If it has no target, it is a statistic. If it has no owner, it is decoration. If it is not associated with a business objective, it is noise with decimal places.

KPI, metric and indicator: the distinction that matters

Not every metric is a KPI. A factory can measure hundreds of things — the temperature of a dyeing oven, electricity consumption per shift, machine pauses. These are operational metrics. They become KPIs when someone decides that one of them is key to an objective and assigns it a target. The discipline is not in measuring more; it is in choosing the few that move the business and ignoring the rest with courage.

There is also an intermediate family that confuses everyone: support indicators, or drivers. Electricity consumption per shift is not a management KPI, but it is a driver of the transformation cost, which in turn feeds margin — that, yes, is a KPI. Understanding this chain — from the operational driver to the management KPI — is what allows you to explain why a top-level indicator moved. A dashboard that only shows the top-level numbers is pretty and useless when someone asks «why».

The anatomy of a defensible KPI

A KPI that survives a management audit has six components:

• Unambiguous definition — «OEE» is not enough; you have to define availability, performance and quality and the sources of each factor.
• Documented calculation formula — whoever calculates it must always arrive at the same number from the same data.
• Single, traceable data source — the number comes from the ERP, the MES, the POS, not from a parallel sheet.
• Update frequency — daily, per shift, real-time; the frequency has to match the speed of the decision.
• Target and alarm threshold — the value to reach and the value at which someone has to call someone.
• Named owner — one person, not a department.

Apply these six to a concrete case. A defensible «margin per order» KPI defines margin as sale price minus real transformation cost minus raw material consumed at the order's rate; it documents that the transformation cost comes from the hours recorded on the shop-floor terminals multiplied by the cost centre's hourly rate; it pulls everything from the ERP with no intermediate sheets; it updates at the close of each order; it triggers an alarm below a target margin defined per family; and it has the sales director as its owner. Leave out any one of these six pieces and the number loses its defence in the meeting — someone will always be able to say «that figure is not right» and be correct.

Lagging indicators and leading indicators

The most useful distinction that the management literature has left us, and the most ignored in SMEs, is between lagging indicators and leading indicators. Last month's margin is lagging — it measures the result after everything is done. The confirmed order book, the capacity utilisation rate for the next four weeks, the number of samples approved per buyer, are leading — they tell you what is going to happen in time to change it. Portuguese SMEs collect lagging indicators with devotion and build almost no leading ones, which explains why they always manage by reacting and never by anticipating.

A brief history of how we got here

The modern KPI concept inherits from two lines: the management accounting of the early twentieth century (the financial metrics) and the quality and lean production movement of post-war Japan, which brought OEE and the obsession with visible waste. In the 1990s, Kaplan and Norton's Balanced Scorecard tried to marry the two families — financial, customer, internal processes, learning. Most Portuguese SMEs ended up somewhere in the middle: they adopted the financial vocabulary and ignored the operational part, because that required data the systems could not provide in useful time.

The historical irony is that OEE was born on the Japanese shop floor precisely because the data had to be next to the machine, visible, updated to the minute, to serve whoever was operating. Decades later, many Portuguese factories have OEE on a sheet the office produces once a month — inverting exactly the logic that gave rise to the indicator. The operational KPI that lives far from the operation is a contradiction in terms.

3. The landscape in Portugal today

Here the numbers matter, and they are uncomfortable. In 2025, according to the INE, only 45% of companies in Portugal carried out data analysis (big data/analytics) — up 6.4 points from 2023, but still far from universal. In other words: more than half of Portuguese companies still manage without a structured practice of data analysis.

The picture becomes clearer when you look at the foundations. Still in 2025, the INE indicates that only 53.7% of companies used an ERP and 38.7% purchased cloud computing services. Without an integrated core, BI works on scattered data — and that is exactly why so many dashboard projects end up feeding on spreadsheets exported by hand.

You cannot build a reliable KPI dashboard on a foundation of manual exports. The data has to be born in the system where the operation happens.

What the INE numbers reveal about maturity

Cross-referencing these three data points gives a brutal diagnosis: almost half of companies want to analyse data, but almost half do not even have an ERP feeding that data in an integrated way. There is an execution gap here. The will to manage by data exists; the infrastructure to do it reliably, very often, does not.

This gap widens by size. In large companies, ERP adoption and data analysis approach universal; in micro-enterprises — which are the overwhelming majority of the Portuguese business fabric — it plummets. The problem is that the national average hides this bimodality: there is a digitalised Portugal and a Portugal that still manages in a notebook, and the two realities coexist in the same cluster, sometimes on the same industrial street in Famalicão. An SME of 60 people that still does its costing in Excel is not behind the average — it is behind its direct competitors that no longer do so.

European pressure and the skills deficit

The European Union's Digital Decade Compass set targets for SME digitalisation by 2030 that Portugal, at its current pace, will struggle to meet. But the problem is not just about systems — it is about people who know how to use them. The shortage of digital skills is a cross-cutting brake: what use is setting up a BI layer if nobody in the company knows how to put an associative question to the data? Technology without skill generates a second silo — a sophisticated dashboard that only one person can operate, re-enacting the «IT hero» problem in a more expensive format.

Industrial sectors move at different rhythms

The Portuguese metallurgical and metalworking sector — more than 23,000 companies and around 250,000 people employed, according to AIMMAP — is living an Industry 4.0 transition in the Aveiro–Marinha Grande corridor, with moulds and injection where the single piece and the large series coexist in the same shed. Here the KPI gets complicated: the costing of a single mould — months of work, prototypes, corrections — has little to do with the costing of an injection series of millions of pieces, and a dashboard that treats the two with the same ruler misleads on both.

The textiles of the Vale do Ave face a double pressure: fast fashion on one side, batch traceability requirements from the EU Strategy for Sustainable and Circular Textiles on the other. The digital product passport that Europe is preparing will require tracing each batch — fibre origin, finishing processes, dyeing — in a way that only an ERP with serious batch traceability can provide. The company that today does not record the batch in the system will discover, within a few years, that traceability has ceased to be an extra and has become a condition of access to the European customer.

The footwear of Felgueiras and S. João da Madeira bets everything on collections with thousands of SKUs and rigid buyer windows. APICCAPS documents a heavily export-oriented sector, where the delivery date to the international buyer is non-negotiable — arrive late at the fair, and you lose the entire season. Here the KPI that matters is not one of production, it is one of plan fulfilment against the buyer's calendar.

Commerce as a thermometer

On the side of retail and distribution, the turnover of commerce in Portugal reached €201.8 billion in 2024, a growth of 3.6% over 2023, with retail commerce growing 4.7% (INE). Growth with tight margins means one thing: whoever does not measure margin per SKU, per store and per channel in near real time is managing blind in a market that moves.

Regional food retail is the most acute case. A chain of 20 stores with single-digit gross margins cannot afford to discover at month-end that a store is selling below cost due to a labelling error, or that an entire category has slipped in shrinkage. The certified POS that communicates with the backoffice in near real time turns each sale into an immediate data point — but only if the backoffice has someone to read that data and act on it before the close.

Financing: what the support pays for and what it leaves owing

There is a real financing engine behind this race — PT2030, PRR, COMPETE 2030, Norte 2030. Many digitalisation and management systems projects are eligible, and IAPMEI supports a substantial part of these applications. Field experience, however, is sober: what usually gets stuck is not the approval of the application — it is the final technical report, the demonstration that the investment generated the promised result. Companies that bought systems without first defining which KPIs would improve arrive at this phase with no way of proving the impact. Defining the indicators before the project is not just good management — it is the documentary proof the application will demand at the end.

4. The KPI implementation models

There are four practical approaches to getting KPIs working in a Portuguese industrial SME. They are not mutually exclusive — most companies evolve from one to the next — but each has its own cost and its own reliability ceiling.

Model 1: spreadsheets — the comfortable swamp

Every company starts here, and there is no shame in that. The spreadsheet is unbeatable for prototyping a new indicator. The problem appears at scale: when three people maintain three versions of the «official sheet», the number ceases to be reliable and meetings start to argue about whose number is right instead of what to do with it. Rule of thumb: if two departments present the same KPI with different values in the same meeting, they left the spreadsheet phase a long time ago.

There is a second risk in the spreadsheet that is rarely admitted: the fragility of the formula. A cell copied with the wrong range, an inserted row that the sum does not catch, a filter left applied from the last analysis — and the number comes out wrong without anyone noticing, because the sheet does not throw an error, it gives a plausible number. The worst data failure is not the one that blows up; it is the one that goes unnoticed and feeds a decision. Informal audits of SME costing sheets frequently reveal formula errors that distort margins by whole percentage points — and that have been guiding pricing decisions for months.

Model 2: native ERP reports

A well-parameterised vertical ERP provides reliable daily listings on invoicing, purchasing, stock and production. It is an enormous leap from the spreadsheet because the data is single-source. The limitation is rigidity: cross-referencing margin per customer with delivery punctuality and defect rate, all on the same screen, requires a flexibility that native reports rarely have.

Do not underestimate this model, however. For a small company with simple processes, well-designed native reports read with discipline resolve 80% of the need at almost zero cost. The error is getting stuck in this model when complexity grows — when it starts to be necessary to answer questions the report programmer did not foresee, and each new question requires a development request and weeks of waiting. At that point, the cost of maintaining the model has already exceeded the cost of abandoning it.

Model 3: BI layer over the ERP

This is where most industrial SMEs should be. A BI tool like Qlik Sense reads the ERP and allows associative analysis — clicking on a customer and instantly seeing all of that customer's orders, margins, delays and returns without building a new report. It uses OLAP technology to answer questions nobody foresaw when the system was set up. It is the difference between having answers and being able to ask questions.

The associative advantage is hard to explain without seeing it, but the test is this: the sales director notices that an old customer has fallen in billing. In a reporting model, he asks IT for a report on that customer's trajectory and waits. In an associative model, he clicks on the customer and sees, in the same second, that the drop coincides with a rise in returns of a specific product family — and realises it is not the customer cooling off, it is a quality problem driving them away. The question that generates the valuable answer is almost always the second or the third, and only tools that let you chain questions in real time make them possible.

Model 4: BI plus real-time shop floor

For operational KPIs — OEE, scrap, setup time — monthly data is useless. It has to be captured at the moment the machine produces it. A production capture system like KORA Productivity collects from the industrial terminal in real time, and the result is that the supervisor sees the shift's OEE while the shift is still under way — and can act. We delved into this mechanism in KORA Productivity: real-time shop-floor control.

The critical point of this model is not the technology — it is the discipline of field collection. An industrial terminal that the operator has to feed with the reason for each stoppage only works if the collection is so simple that it does not steal time from production, and if the operator sees a return on having reported. Rollouts that impose recording bureaucracy without giving useful information back to the shop floor die within weeks: the operator learns to always press the same button to get rid of the screen, and the data that reaches the dashboard is rubbish with the appearance of rigour. The golden rule: whoever feeds the data has to be the first to benefit from it.

A financial KPI can be monthly. An operational KPI has to be of the shift, or it is no use for managing the shift.

5. How to assess whether your company needs it

You do not need a six-month study to know where you stand. There are clear signs, and there is a quick method to read them.

Symptoms that your KPIs are not working

• Management meetings spend more time validating numbers than deciding based on them.
• The key report is produced by one person and, if that person is absent, nobody can reproduce it.
• Two departments present different values for the same indicator.
• The numbers arrive too late to change anything at all.
• Nobody can explain how a KPI is calculated without consulting a hidden spreadsheet.
• When the KPI turns red, there is no named person who has to act.
• The company measures many things but cannot say which five numbers matter most.

Step by step: a five-stage diagnosis

1. List the five indicators management looks at first. If it takes more than ten minutes to arrive at five, the problem is not technical — it is one of focus.
2. For each, identify the data source and the real update frequency. Write next to it whether it comes from the system or from a manual sheet. The number of «manuals» is your integration deficit.
3. Measure the decision latency. How much time passes between the fact happening on the shop floor and the KPI reflecting it? Compare that latency with the speed at which you need to decide.
4. Name the owner of each KPI. If you cannot name a person (not a department), the indicator has no owner and nobody acts when it worsens.
5. Classify each KPI as financial or operational. The financial ones tolerate daily updating; the operational ones need real time. This simple ordering tells you where to invest first.

Reading the diagnosis result

The patterns this diagnosis reveals are recognisable. If most KPIs come from manual sheets, the problem is one of data source — resolve integration first, before dreaming of dashboards. If the KPIs come from the system but the latency is enormous, the problem is one of collection frequency — shop-floor capture is probably missing. If the KPIs are reliable and fast but nobody acts when they flash red, the problem is one of governance — the owner and the escalation rule are missing. Each pattern points to a different investment, and spending on the wrong problem is like buying new tyres for a car with no engine.

The decision matrix: where to invest first

Quick wins for the next two weeks

Before any large project, three things pay off almost immediately: (1) choosing one critical operational KPI and moving it from monthly to daily, even if only on a whiteboard on the shop floor; (2) killing two or three KPIs that nobody uses but that are still calculated out of inertia; (3) forcing a single, written definition of the indicator that most generates argument in meetings.

The shop-floor whiteboard deserves a defence: it is the opposite of sophisticated and is frequently the step with the best return per euro invested in this whole journey. A number updated by hand every shift, visible to those who work there, changes behaviours in a way that no dashboard hidden in the office can. Only after the culture of looking at the number exists does automating that number bring value. Automating an indicator nobody consults is speeding up the production of something nobody wants.

6. What to choose and why — decision by company size

The right recommendation depends on size and complexity, not fashion. A company should not buy the data architecture of another three times its size.

Fewer than 30 employees: resolve the source before anything

At this size, the most common sin is wanting BI before having data. The absolute priority is to have invoicing, purchasing, stock and — if there is production — work orders being born in a single, certified system. A vertical ERP parameterised for the sector, with native reports read with discipline, covers the overwhelming majority of the need. The spreadsheet can continue to exist for prototyping new indicators — but never as the official source of a management KPI. Investing in sophisticated dashboards at this stage is decorating a house before it has foundations.

30 to 100 employees: the operational turning point

It is in this bracket that most Portuguese garment factories, footwear factories and metallurgical units find themselves, and it is here that shop-floor capture moves from luxury to necessity. With significant production, operational KPIs — OEE, scrap, plan fulfilment — only have value if they are of the shift, and that requires terminals in the field. At the same time, the first BI dashboards over the financial KPIs start to make sense. The order matters: first the operational data in real time, then the analytical layer that cross-references it with the financial.

100 to 250 employees: aligning who measures and who decides

Here the problem is no longer having data — it is reconciling data from departments that grow in silos. It is the bracket where the cold war between finance and operations breaks out most fiercely, and where an associative BI layer with serious data governance becomes a condition of sanity. Governance — who defines, who validates, who alters — ceases to be optional: without it, each area builds its own dashboard with its own definitions, and you return to the problem of numbers that don't reconcile, now with more expensive technology.

The trap of buying above your weight

An error we have seen for decades: the company of 40 people that buys the data architecture of one of 400 because «it is for growing». The predictable result is a system nobody has the structure to feed, empty dashboards and a sense of failure that sets back digital maturity by years. The rule is the opposite: resolve the reliability of the base data first, then add the analytical layer when the data is trustworthy.

Buying the data architecture of a company three times larger does not make you grow faster — it makes you inherit problems you don't yet have.

The case of extreme complexity

For companies with very complex manufacturing processes — multiple subsidiaries, dense business rules, heavy integrations — a low-code ERP like QAD Adaptive allows you to model what generalist ERPs cannot. And when there is a need to link operations between subsidiaries and partners, integration via Multi Connect avoids the proliferation of reconciliation sheets between units. On the pattern of automating what is today manual, Process automation in Portuguese industry: an operational guide is worth reading.

KPIs by type of operation

In a warehouse management operation, the reigning KPIs are inventory accuracy, picking productivity and dispatch window fulfilment — and poorly calculated safety stock is the silent cause of half of all logistics fires. In omnichannel retail, indicators such as click-and-collect conversion rate and margin per channel come into play. Each operation has its half-dozen indicators that matter; the sin is copying another's list.

KPIs by sector: the half-dozen that matter

7. Regulatory framework and applicable compliance

KPIs do not live in a legal vacuum. The data that feeds them is subject to rules that, ignored, turn a BI project into a risk exposure.

Invoicing, SAF-T and the fiscal source of the data

Many financial KPIs are born from invoicing documents. DL 28/2019 and the mandatory nature of ATCUD and of software certified by the AT mean that the source of these indicators has to be a certified system — not a spreadsheet. The monthly SAF-T communication (Ordinance 195/2020) creates, moreover, a clean and structured dataset that is gold for analysis, if it is put to use.

It is worth underlining a little-explored practical consequence: the SAF-T that the company is already required to produce contains, in a standardised format, a substantial part of the data that a financial KPI dashboard needs. The company that treats SAF-T merely as an obligation to fulfil is throwing away an analytical asset it has already paid to build. The same discipline that ensures fiscal compliance ensures the reliability of the data source — the two things are resolved together.

Data protection in people KPIs

As soon as you measure HR indicators — absenteeism, turnover, individual productivity — the GDPR and Law 58/2019 come into play. People indicators require a legal basis, minimisation and special care with any analysis that approaches profiling. A people management tool with predictive AI for turnover has to be designed with these limits in mind — and, depending on the use, under the eye of the AI Act (EU Regulation 2024/1689).

The line that separates the admissible from the problematic is subtle. Measuring the aggregate absenteeism of a section is legitimate management; using an algorithm to assign each employee a departure risk score that influences decisions about that person enters profiling territory that the GDPR and the AI Act surround with heightened requirements — transparency, robust legal basis, the right to human intervention. The rule of thumb: the more individual and the more consequential the people indicator, the more legal rigour it demands. And companies with 50 or more employees also have the obligation of the whistleblowing channel under Law 93/2021 to consider when designing the governance of their HR data.

Cybersecurity of the indicator repository

A management dashboard concentrates the most sensitive things the company has: margins, customers, costs, salaries. It is a target. Security monitoring and the requirements of digital operational resilience and of NIS2 (EU Directive 2022/2555, transposed by DL 65/2025) make the protection of this repository a matter of compliance, not just of prudence. Certifications such as ISO 27001 provide the framework; the cybersecurity service provides the execution.

NIS2 substantially broadened the universe of entities covered compared with the previous directive, catching sectors and company sizes that were previously left out — including many medium-sized industrial companies that considered themselves too small for the regulatory radar. The common error is to assume that the cybersecurity of the KPI dashboard is a problem for the IT department. It is not: a leak of margins per customer to a competitor, or a malicious alteration of the numbers that guides a wrong decision, are top-level business risks. The indicator repository deserves the same level of protection as the customer database.

The KPI dashboard is the company's safe in digital form. Treat it with the same respect with which you treat the physical safe.

8. How INFOS approaches this

Thirty-six years building software for Portuguese industry have taught us something the generalist vendors do not know: a KPI is only reliable if it is born in the system where the operation actually happens. That is why our approach starts with the source data — in the vertical ERP parameterised for textiles, footwear, apparel, metal or plastic — and only then rises to the analytical layer.

In practice, this means capturing production on the shop floor in real time with industrial terminals, ensuring that invoicing comes from a certified system, and connecting everything to an associative BI layer where the director asks questions instead of waiting for reports. When needed, we add AI components for forecasting — always within the regulatory limits that the analysis of people and business data imposes.

Why verticalisation changes costing

The detail that generalist ERPs do not resolve is product modelling. A footwear collection with three axes — colour, size, last — and 1,200 SKUs does not fit into a system designed to sell screws. If the ERP does not correctly model the product structure, the costing comes out wrong at the source, and every margin KPI built on top inherits the error. Verticalisation is not a question of comfort of use — it is the condition for the margin per reference to be a number you can trust. What we learned in thousands of projects serves us here: the most beautiful KPI always rests on an invisible product parameterisation that nobody on the dashboard sees.

What we do not do

We do not sell beautiful dashboards over bad data. We sell the discipline of having a single, reliable data source, and the layer that turns it into a decision. It is less glamorous than a demo with moving charts, and it is what separates a project that lasts from one that fills the hard drive with screens nobody opens. Nor do we hide that we have seen projects derail — almost always from the same cause: starting with the visualisation layer before resolving the source, or imposing shop-floor collection without giving value back to those who collect. We have learned to start small, prove value early, and expand only when the data is trustworthy.

9. 30/60/90 day roadmap

A KPI project is not done in a big bang. It is done in layers that prove value early, because the warehouse manager who leaves the radio for more than two hours for a training session will sabotage the rollout — and he is right to do so if he does not see a quick return.

Days 1–30: foundation and focus

• Run the five-stage diagnosis from section 5 and produce the list of the KPIs that really matter.
• Kill the dead indicators — those calculated out of inertia and that nobody uses.
• Write the single definition of each surviving KPI: formula, source, frequency, target, owner.
• Identify which are born from reliable system data and which depend on manual sheets.

The concrete deliverable of this phase is no software at all — it is a one-page document per KPI. If the company cannot produce that document, it is not ready to automate anything. And frequently this phase, on its own, already improves management: forcing the writing of the definition of an indicator that generated argument resolves the argument, because it exposes that the two sides were calculating different things with the same name.

Days 31–60: single data source and first dashboards

• Eliminate the manual sources of the three most critical KPIs, linking them directly to the ERP.
• Build the first BI dashboard over those three, with daily updating at a minimum.
• If you have significant production, start capturing an operational KPI (OEE or scrap) in real time at a pilot station.
• Officially name the owners and define what happens when each KPI turns red.

Choose the pilot station carefully: not the most problematic — the most cooperative. The objective of this phase is to prove that real-time collection generates information that changes the shift, and that needs a supervisor who embraces the change, not one who resists it. A successful pilot at a friendly station becomes the argument that convinces the rest; a failed pilot at the most difficult station buries the entire project.

Days 61–90: expansion and governance

• Extend the real-time capture from the pilot station to the line or section.
• Introduce cross-analysis — margin against punctuality against defect — into the same dashboard.
• Formalise data governance: who can alter definitions, who validates, how often the KPI list is reviewed.
• Carry out the first quarterly review: which KPIs changed real behaviours? Those that changed nothing

Frequently asked questions

What is an operational KPI and how does it differ from a financial indicator?

An operational KPI measures shop-floor performance in real time — deadline fulfilment, defect rate, machine efficiency. A financial indicator measures a later result — margin, stock turnover, DSO. The Portuguese problem is that the operational ones arrive late and the financial ones do not explain why the factory worked well but the company lost money.

Why do KPI reports in Portuguese industrial SMEs always arrive late?

Because they are built manually in Excel, cross-referencing data from various silos — production sheets, transport notes, work orders. One person dedicates two days a month to pasting cells. By the time the data is ready, the decisions have already been made. The architecture is the problem, not incompetence. Automating is the solution.

How do you know if a KPI is measuring the wrong thing?

Do the test: if this number doubled tomorrow, would the company be better off? If the answer is «it depends», you are measuring effort, not value. Pairs produced per day can double with errors. Metres per shift can double with colour deviations. A good KPI bothers someone because it exposes real problems.

What is the biggest risk of a KPI that rises consistently?

It is almost always the wrong indicator. A number that grows non-stop frequently hides a metric that measures activity instead of outcome. In footwear, pairs/day ignores the fulfilment of critical dates. In textiles, metres/shift ignores shade deviations. Consistency is suspect.

How do you resolve the conflict between the CFO and the COO over which KPI to use?

It is not about choosing one. It is about creating a third indicator that reconciles both — for example, margin generated per machine-hour instead of just OEE. Without that link, each area optimises its own number and the company as a whole loses. Alignment requires an indicator both recognise.

What is the hidden cost of a manually built KPI?

The opportunity cost of the person who builds it. In an SME of 80 employees, it is frequently the best knower of the business spending two days a month in Excel. When that person leaves, the KPI disappears with them. You did not have a process — you had a person. That is the greatest risk.

How long does it take to implement an operational KPI system in a Portuguese factory?

The guide proposes a 90-day plan that any company in textiles, footwear, distribution or metal can execute without stopping production. It is not a two-year BI project. It is operational before it is technical — it starts by aligning what is measured, then automates the collection.

Sources

• Instituto Nacional de Estatística (INE) — Industrial Production Statistics and Economic Activity Indicators in Portugal
• ISO/IEC 27004:2016 — Information security management — Measuring information security
• IAPMEI (Institute for Support to Small and Medium-Sized Enterprises and Innovation) — Operational Management Guides for Industrial SMEs
• Banco de Portugal — Economic Situation Reports and Business Performance Indicators
• ISO 22400:2018 — Automation systems and integration — Key performance indicators (KPIs) for manufacturing operations management