Most Portuguese industrial companies measure too much and understand too little. The shop floor generates numbers. The office collects them. No one knows for sure which of them matters. And here is the real problem: according to INE, only 45% of companies in Portugal carried out data analysis in 2025. The other half collects noise.
We see this repeatedly. A textile factory in the Vale do Ave with 120 employees has 47 charts on a Qlik Sense dashboard. The production manager opens it at 7.30 a.m., sees a red line, swipes to the next. No action. The numbers exist, but they do not speak. The problem is not a lack of data — it is a lack of judgement. And a lack of judgement kills the investment.
The invisible cost of a metric without purpose
Every KPI you measure has a cost. Not in money — in attention. Attention is the rarest asset on a production line or in a warehouse. When an operator or a shift supervisor has to report 15 daily metrics, they end up reporting none of them accurately. They fill in the numbers the computer expects. The data is corrupted at source.
Five years ago, we ourselves believed that more visibility was always better. The more KPIs, the more control. We were wrong. What we see today in successful implementations is the opposite: teams that measure few indicators but that act on every one of them. The difference is enormous.
A KPI that does not lead to an action is just noise with pretensions of intelligence.
Take a concrete example. A garment manufacturer in Famalicão implemented a dashboard with OEE (Overall Equipment Effectiveness), WIP (Work in Progress), absenteeism rate and cost of defects. Four numbers. Nothing more. Each of them had an owner — the shift supervisor knew that metric X was their responsibility. When OEE dropped 2%, there was a 15-minute meeting. Cause identified, action taken. Result: OEE went from 61% to 74% in five months. By comparison, a neighbouring garment manufacturer with 23 KPIs on an "advanced" BI dashboard kept OEE at 58% and no one knew why.
The second cost, less visible, is confusion. When there are too many numbers, priority disappears. The manager sees 47 charts and asks: "Which one do I start with?" The answer is: none of them. Because the dashboard was designed to look complete, not to look useful.
What to measure: a survival criterion
There is no universal set of industrial KPIs. A footwear factory does not measure the way a textile one does. A distributor does not measure the way a garment manufacturer does. But there is a criterion that separates what matters from what merely looks good on a slide.
A KPI deserves to exist if it answers one of these three questions: does it affect direct profitability (OEE, scrap rate, labour cost per unit, gross margin per product)? Does it affect the ability to meet deadlines (cycle time, delay rate, WIP, supplier lead time)? Does it affect operational risk that can halt the business (safety, critical quality, availability of critical stock)?
If a number does not answer any of these, do not measure it. Delete it from the dashboard. No one needs to know the occupancy rate of the meeting room or the number of emails answered per day — unless the business depends on them (and it rarely does).
Take a construction materials distributor with 35 employees and 4 warehouses. The KPIs that matter are: average days of stock (affects cash flow and space), picking time (affects meeting deadlines), customer returns (affects margin and reputation), transport cost per order (affects profitability). Everything else — number of pallets moved, average handling time, forklift utilisation rate — is detail. And yes, there are companies that measure detail, because someone thought it was clever to put a sensor on every machine. Then they end up with 200 numbers and no decisions.
What to ignore (and why most do not)
The temptation is to measure everything. IIoT technologies have made sensors cheaper. Qlik Sense dashboards let you visualise anything in seconds. The result is that many industrial companies accumulate data the way a warehouse accumulates stock — without judgement, hoping that one day someone will find value in it. Ignoring takes courage. Because someone will ask: "Why don't we have this KPI?" And the honest answer is: "Because no one is going to use it to decide anything." But that answer is uncomfortable. It is easier to include the number.
Vanity metrics are the first culprit. "Number of units produced per day" looks important on a dashboard. But if the margin is nil, producing more is producing losses. What matters is profitability, not volume. Then come orphan metrics — those that no one is responsible for driving up or down. If no one owns it, it is an ornament. Delete it. Then duplicate metrics: "Deadline compliance rate" and "Percentage of orders on time" are the same thing. Choose one, delete the other.
A case we saw: a plastic injection moulding company measured "number of downtime hours per machine and shift". It looked clever — it identifies problematic machines. Problem: the data was filled in manually by the operators, who rounded it. The KPI was an estimate. Result: preventive maintenance decisions based on estimates. The company spent 40 thousand euros on unnecessary replacement parts before realising the numbers were fiction. The lesson is hard: a KPI measured by accident (because there is a field in the ERP) is worse than no KPI at all.
From measurement to action: the missing link
The biggest failing of industrial dashboards is not the lack of data. It is the silence after the chart appears. A well-designed dashboard has three components: the metric (the number), the alert (when it drops below X, it turns red) and the action (who does what, when it turns red). If there is no action, the dashboard is just a pretty monitor.
KPI: OEE < 65%. Owner: Production supervisor. Action: 15-minute meeting with the shift team. Root cause. Adjustment decision. Deadline: Same week.
Without this, the dashboard is an exercise in data collection, not in management. We see many BI implementations that fail because the company measures well but does not decide well. The KPI appears, no one knows who is responsible for the action, a week goes by, no one remembered, the problem carried on. A month from now, the dashboard is ignored. The investment in data was wasted.
An example of a minimalist dashboard that works
A garment manufacturer in the Vale do Ave with 85 employees implemented a dashboard with just five KPIs: daily OEE (production line), absenteeism rate (capacity forecast), critical supplier lead time (risk of disruption), cost of defects (quality + profitability), days of raw material stock (cash flow). Each KPI had a named owner. Each had an acceptable range (e.g. OEE between 70-80%, absenteeism < 5%). When one went out of range, there was action the same day.
Result: in 18 months, OEE rose from 62% to 77%, absenteeism fell from 8% to 4%, cost of defects dropped 35%. The dashboard was not pretty — it was black and white, with no effects. But it was used. By comparison, another garment manufacturer in the region with an "advanced" dashboard of 31 KPIs and 3D visuals, implemented by the same BI consultancy, had a 60% abandonment rate. The manager opened it once a week, saw numbers, did not know what to do, closed it. A year later, the dashboard was switched off. The investment was lost because no one defined what to do with each number.
The role of data integration
A KPI is only as good as the quality of the data that underpins it. And the quality of the data depends on integration.
If your company has a well-configured MULTI ERP or QAD Adaptive, with Multi Connect integrating subsidiaries or partners, the data reaches the dashboard clean. If you have data scattered across multiple systems — a generic ERP, a production spreadsheet, a quality file on the shift supervisor's computer — then you are feeding the dashboard with fiction. According to INE, only 53.7% of companies in Portugal used an ERP in 2025. That means almost half are trying to do BI without an integrated core. It is like trying to navigate with a map made of pieces of different maps.
Integration is not a luxury. It is the foundation. Without it, every KPI is an estimate. And estimates lead to wrong decisions.
In summary: fewer numbers, more action
The most common mistake in industrial BI implementations in Portugal is the confusion between "having a lot of data" and "having intelligence". Data without judgement is just amplified noise. A dashboard with 47 charts and no owner is a monument to waste. A dashboard with 5 KPIs, each with an owner and an action plan, is a management tool.
If you are designing a dashboard, start by answering this: what is the action that each number will trigger? If you cannot answer, delete the number. Then, name an owner. Then, define the acceptable range and what to do when it goes out. Then, implement it. The rest is decoration.
Frequently asked questions
How many KPIs should an industrial Qlik Sense dashboard have?
There is no magic number, but practice shows that between 4 and 8 KPIs per dashboard is ideal. Each should have a responsible owner and an associated action. Dashboards with more than 15 indicators tend to generate inaction — the user does not know where to start. Quality over quantity.
What is the difference between a vanity KPI and a useful KPI?
A vanity KPI looks important on a slide but does not lead to decisions. "Number of units produced" is vanity if the margin is nil. A useful KPI answers three criteria: it affects direct profitability, the ability to meet deadlines, or operational risk. If it answers none of them, it is an ornament.
How do you know if a KPI should be removed from the dashboard?
Remove a KPI if: no one is responsible for it (orphan metric), it duplicates another indicator, or it does not lead to action when it changes. If a number exists only because there is a field in the ERP, it is worse than not existing. Data filled in without purpose corrupts decisions.
Is OEE really the best KPI for factories?
OEE is excellent if the company acts on it. A garment manufacturer that monitors OEE with a responsible owner and weekly action went from 61% to 74% in five months. But OEE in isolation, without the context of profitability or deadlines, is just a number. Combine it with cost of defects and WIP.
Why do companies with lots of IIoT sensors have worse results?
Cheap sensors generate data in excess. Without a selection criterion, the company accumulates 200 numbers and no decisions. The problem is not the lack of data — it is the lack of attention. Operators with 15 daily metrics to report fill in the numbers the system expects, corrupting the source.
What is the real cost of measuring a metric without purpose?
The cost is in attention, the rarest asset in production. Every KPI you measure consumes the time of operators, shift supervisors and managers. When there are too many numbers, priority disappears and no one acts on anything. A KPI without action is noise with pretensions of intelligence.
How do you structure a KPI to ensure there is action?
Every KPI should have three components: the metric (the number), the alert (when it turns red), and the action (who does what). Example: OEE below 65% triggers a 15-minute meeting with the shift team to identify the root cause. Without a defined action, the dashboard is just a pretty monitor.