Forty-five per cent of Portuguese companies already carry out data analysis — and most continue to decide on intuition. Not because the data is lacking. Because no one has defined which decisions should be fed by data, how often, and who answers when the number diverges from the supervisor's experience. That gap is not resolved with a new dashboard. It is resolved with a decision architecture that most companies still do not have — and that this article describes, step by step, without slide-deck optimism.
What you need before you begin
A data-driven culture is not born from a tool. It is born from conditions that many companies still do not have in place. According to INE (2025), only 53.7% of companies in Portugal used ERP — which means almost half of the Portuguese business fabric still does not even have the integrated core that makes any analysis reliable. Before proceeding, check whether these foundations exist in your operation.
You need a central system that records transactions in real time — not in an overnight batch. You need data with an owner: each indicator has a person responsible for its quality, not a department. You need at least one middle manager willing to be contradicted by a number — and who does not interpret that contradiction as an affront to their experience. You need comparable history with at least twelve months of depth, minimum integration between systems (without an API or synchronisation file, BI works on dispersed data), an agreed definition of at least five operational KPIs, and a management meeting with a fixed cadence where the data is presented and debated — not sent by e-mail the day before and ignored in the room.
If three or more of these points are missing, the problem is not one of tooling. It is one of information architecture. Solve that first.
Step 1 — Define what "deciding with data" means in your operation
Here is the error we see repeated in INFOS projects: the company buys a BI and dashboards tool, installs it, and three months later the dashboard is open on a screen in the meeting room — but decisions continue to be made based on the intuition of the production director. The problem is not the tool. It is that no one defined which decisions should be fed by data and how often.
A dashboard that feeds no specific decision is expensive decoration.
Do this exercise with your management team. List the ten most repeated operational decisions in your company — when to launch a production order, when to replenish stock, when to escalate a delivery delay. For each one, identify: is it made based on data or on experience and intuition? Of those made on intuition, which have data available that is not being used? Choose three. Only three. Define the data that should feed each decision and who takes it.
Do not try to change everything at once. A data-driven culture is built decision by decision, not by decree. What you should verify at the end of this step: the ten decisions are listed, each has a "data" or "intuition" classification, and the three priority ones have a defined data source and owner.
Step 2 — Audit the quality of your data before trusting it
In a typical clothing factory in the north of the country, it is common to find three versions of the same efficiency indicator: one in the ERP, another in the supervisor's spreadsheet, another in the production director's head. When the three diverge, the meeting spends forty minutes discussing which is the right number — and zero minutes deciding what to do. There is a detail that management manuals rarely mention: the first time the dashboard contradicts the supervisor's experience and the supervisor is right, you lose the cultural battle for months. Distrust in data is far harder to reverse than the absence of data.
Audit each source with four questions. Origin: is this data entered manually or captured automatically? Latency: when the event occurs, how long does it take to appear in the system? Consistency: does the same indicator have the same definition across all departments? Completeness: what percentage of records have no null value in the critical fields?
Data entered manually with a twenty-four-hour latency does not support the day's operational decisions. It supports, at most, retrospective analysis. Know what you have before you promise what you can deliver. Document the origin and latency of each critical indicator, write down the definition of each KPI and share it with all users — do not keep it in the head of the IT analyst.
Step 3 — Install the decision cadence
A data-driven culture does not live in dashboards. It lives in meetings. The difference between a company that uses data and one that collects it lies in the cadence with which that data enters the decision cycle.
The dashboard no one opens in the Monday meeting is not a technology problem. It is a ritual problem.
Define three levels of cadence. Daily, ten minutes: production, dispatch and previous day's anomaly indicators. Shop floor or warehouse. No PowerPoint, no elaborate prior preparation — the data is on the screen, it is read together, deviations are identified. Weekly, thirty to forty-five minutes: operational KPIs versus target, deviations with identified cause, corrective action with owner and deadline. Monthly, ninety minutes: trends, comparison with the same period last year, review of targets. CFO, COO and area managers in the same room, over the same numbers.
Each meeting follows a fixed format: data, deviation, cause, action. Four fields. Nothing more. To understand how to align the CFO and COO on the same indicators without the meeting turning into a debate about which version of the number is correct, see the article Industrial KPIs the CFO and COO disagree on — and how to align them.
Step 4 — Choose the right KPIs for your sector
There is no universal list of data-driven KPIs. There is a right list for your business model. A food distributor in the Lousada-Paços corridor needs to measure order fill rate and cost per picking line. A textile factory in the Vale do Ave needs to measure efficiency per section and rejection rate per batch. A footwear company in Felgueiras with eight hundred active references needs to measure rotation by colour-size-last axis — not overall rotation, which hides the models that are dying in the warehouse. These are different worlds, and a generic KPI applied to the wrong context is more dangerous than measuring nothing at all.
Use this matrix to select the KPIs that are worth the effort to measure:
| Candidate KPI | Direct financial impact? | Data available today? | Decision it feeds | Priority |
|---|---|---|---|---|
| OEE per production line | Yes | Partial (manual) | Capacity planning | High |
| On-time delivery rate | Yes | Yes (ERP) | Dispatch management | High |
| Stock rotation per reference | Yes | Yes (ERP) | Replenishment and purchasing | High |
| Customer satisfaction (NPS) | Indirect | No | Commercial strategy | Low (phase 2) |
| Cost of non-quality per batch | Yes | Partial | Production control | Medium |
Fill in this matrix with your candidate KPIs. Eliminate everything that has no direct financial impact and no available data. What remains is your starting point — and it should not exceed seven active indicators in the initial phase. For a more complete guide on selecting industrial KPIs, see Business KPIs: the operational guide for Portuguese industrial directors.
Step 5 — Connect the systems so that data flows without manual intervention
A data-driven culture dies when someone has to export an Excel file on Fridays to feed Monday's dashboard. That someone will be on holiday. Or will forget. Or will export the wrong version of the file — last month's, with this month's file name. It always happens, in every company where we see this process.
Integration between systems is not a luxury. It is the infrastructure of the culture. Map your critical data flows: does the ERP feed the BI automatically or manually, and how often? Does the shop floor record data in real time or at the end of the shift? Does the warehouse record movements at the moment or in batch? Do external sales synchronise in real time or in a daily batch? Each manual flow is a point of failure. Each point of failure is a decision made with wrong data.
Tools such as KORA Productivity capture production data in real time directly from the shop floor, eliminating the deferred recording that contaminates efficiency reports. The KORA Inventory Suite does the same for warehouse movements — the warehouse manager does not step away from the radio, the data reaches the ERP without manual intervention. For textile industries, see how this data translates into actionable dashboards in the article BI for the textile industry: dashboards decision-makers actually use.
The five errors that destroy data-driven projects before they begin
Starting with the dashboard instead of the question. The company buys a BI tool, asks the vendor for "a pretty dashboard" and then asks what to do with it. Reverse it: first define the three decisions you want to improve, then build the dashboard that feeds them. A dashboard without a question is an answer in search of a query.
Measuring everything that can be measured. We see companies with forty active KPIs that act on none of them. The volume of indicators creates the illusion of control without producing any decision. Limit yourself to seven indicators in the initial phase. Add more only when the first ones are stabilised and generating concrete actions.
Ignoring the quality of the source data. A dashboard built on inconsistent data does not create a data-driven culture — it creates distrust in the data. And distrust is far harder to reverse than ignorance. The first time the number contradicts the supervisor's experience and the supervisor is right, you have lost the cultural battle for months.
Not appointing an owner for each indicator. If everyone is responsible for the quality of a piece of data, no one is. Each KPI has an owner. That owner answers for consistency, for latency and for the explanation of deviations at the weekly meeting. No name, no accountability.
Treating the data-driven culture as an IT project. The IT director can install the system. They cannot install the habit of using data to decide. That habit has to be installed by the CEO and the CFO — in the meetings, in the questions they ask, in the decisions they refuse to make without data. When the CEO asks for a number before approving a decision, the whole organisation learns that data matters. When the CEO approves without asking, the whole organisation learns that data is decoration. Consult the glossary on Lean Manufacturing to understand how this logic of continuous improvement based on evidence already exists in other operational methodologies.
Sources
- INE — Survey on the Use of Information and Communication Technologies in Enterprises, 2025
Frequently asked questions
What is the difference between having data and having a data-driven culture?
Having data is having numbers. A data-driven culture is using those numbers to make specific decisions, with a defined cadence and clear accountability. Most Portuguese companies have data but continue to decide on intuition because no one has defined which decisions should be fed by data, how often, and who answers when the number diverges from the manager's experience.
Do I need new BI software to start a data-driven transformation?
No. A data-driven culture is not born from a tool. It is born from conditions: a central system that records transactions in real time, data with an identified owner, managers willing to be contradicted by numbers, twelve months of history, minimum integration between systems, defined KPIs and management meetings with a fixed cadence. If three or more of these points are missing, the problem is one of architecture, not of software.
How do I know whether my data has enough quality to make decisions?
Audit each source with four questions: is the data entered manually or captured automatically? How long does it take to appear in the system after the event? Does the same indicator have the same definition across all departments? What percentage of records are complete in the critical fields? Manual data with a twenty-four-hour latency does not support the day's operational decisions.
Why does a dashboard open in the meeting room not guarantee data-based decisions?
Because a dashboard that feeds no specific decision is expensive decoration. The difference between a company that uses data and one that collects it lies in the cadence with which that data enters the decision cycle — through structured meetings, not ignored screens. The dashboard is only the support; the decision is the ritual.
How many decisions should I try to move to data-driven at the same time?
Only three. A data-driven culture is built decision by decision, not by decree. Start by listing the ten most repeated operational decisions in your company, identify which have data available but unused, and choose three priority ones. For each, define the specific data and the person responsible for the decision.
What should I do when the dashboard contradicts the manager's experience?
If the manager is right and the dashboard is wrong, you lose the cultural battle for months. Distrust in data is far harder to reverse than the absence of data. That is why, before implementing any metric, you should audit the quality of the data and document the origin, latency and definition of each KPI with all users.
How frequent should data-based decision meetings be?
Three levels: daily (ten minutes, production indicators and anomalies), weekly (thirty to forty-five minutes, KPIs versus target and corrective actions) and monthly (ninety minutes, trends and review of targets). Each meeting follows a fixed format: data, deviation, cause, action. No elaborate PowerPoint, no unnecessary prior preparation.