Forty-five per cent of companies in Portugal carry out data analysis in 2025 — and that figure includes those who export one Excel file per month and call it analytics (INE, 2025). The figure that matters most is just below it: only 53.7% of Portuguese companies used an ERP that same year (INE, 2025). This means that almost half of industrial companies are considering self-service BI over scattered sources, with no integrated core. In that context, self-service does not solve the problem — it defers it with a visualisation layer over data that remains inconsistent.
The thesis of this article is this: the greatest obstacle to self-service BI in Portuguese industry is not technological. It is organisational — and it has three concrete faces. The first is the permissions architecture, which almost no one configures correctly at the outset. The second is the quality of the data source, which most factories overestimate. The third is the decision of who "owns" each dashboard, which most companies defer to after installation — and never make. Qlik Sense solves the technical part. What fails is the rest.
The real problem is not the software — it is the data ownership model
Most self-service BI implementations fail in the same way: IT installs the tool, trains two people, and six months later there are forty outdated dashboards that no one trusts. Qlik Sense is not immune to this pattern. What sets it apart is that its associative engine architecture allows a controller or a production supervisor to navigate across dimensions without depending on predefined queries — but that only works if the source data is clean and if someone from the operation has the authority to maintain the model.
Self-service does not mean no governance. It means decentralised governance — and that distinction costs entire projects when it is not spelled out at kick-off.
In factories in the textile sector in the Vale do Ave, we see a recurring pattern: the ERP records production orders, KORA Productivity captures OEE in real time, Qlik Sense aggregates everything — and the line supervisor still asks IT for the week's figures because they were never assigned their own view with appropriate permissions. The software does the work. The organisational structure does not.
Who is the "data owner" in the typical Portuguese factory?
In the Portuguese industrial SME — family-owned, CEO + CFO + IT as a decision-making trio — IT is often one person with 15 years of business knowledge and zero time available to be a data curator. Assigning them ownership of all Qlik datasets is a guarantee that self-service never gets off the ground.
The solution requires three distinct roles defined before any configuration. The operational Data Steward is someone from production or the warehouse who validates whether the figures make sense on the ground — it is not IT, it is the foreman who knows that machine 7 was stopped on Tuesday and that this data has to appear in the dashboard. The Data Engineer maintains the flows from the ERP to Qlik; it may be IT, but with a limited and documented scope. The Dashboard Owner is the business user who publishes and updates the view for their area without needing to program. Without this separation, self-service collapses into centralised dependency — exactly what you wanted to avoid.
What Qlik Sense solves technically that others do not solve in the same way
Associative engine vs. query-based: the operational difference
Most industrial users do not know what a query is. They know they want to see "overdue orders by customer, filtered by last week, with the associated raw material cost". In a query-based tool, that requires someone to have predefined exactly that combination. In Qlik, the associative engine allows the user to perform that navigation in real time, with no new query, no ticket to IT.
For a controller in a footwear factory in Felgueiras — where a collection can have between 800 and 1,200 SKUs with three axes (colour, size, last) — this is not a convenience. It is the difference between being able or not to carry out the analysis before the meeting with the international buyer, who visits twice a year and does not wait.
Integration with industrial sources: where the real work begins
Qlik Sense connects to multiple sources simultaneously: ERP MULTI, shop-floor CSV files, SQL databases, REST APIs. But the integration is not automatic — it requires mapping. The native connectors cover the standard cases; for specific industrial sources such as machine data via OPC-UA, the ETL work has to be done upstream, usually by KORA Productivity, which already aggregates production data into a structured format.
Qlik does not turn dirty data into clean insights. It amplifies what exists — for better and for worse.
Security and access control: what GDPR and NIS2 imply
Self-service BI in an industrial environment is not regulatorily neutral. If dashboards include individual productivity data — pieces per operator, attendance, breaks — we are in personal data territory under the GDPR and Law 58/2019. Qlik has Row-Level Security that allows each user to see only the data for their area, but this configuration has to be done deliberately. It is not the default. Whoever installs Qlik, publishes dashboards with individual data and does not configure RLS is in breach — regardless of whether the software technically allows it.
The second vector is the NIS2 Directive, transposed by DL 65/2025: for companies in critical sectors or that are suppliers to essential entities, access to operational data via browser raises network segmentation questions. A VPN is not sufficient if the architecture is not segmented. Check whether Qlik Sense is deployed in an isolated environment or whether it shares infrastructure with process control systems.
Technical deployment options: a real comparison
| Mode | Initial cost | Time to production | IT dependency | Scalability | Suitability for PT industrial SME |
|---|---|---|---|---|---|
| Qlik Sense SaaS (cloud) | Low | 2–4 weeks | Minimal (managed infrastructure) | High | Good for fewer than 5 sites, non-critical data |
| Qlik Sense Enterprise on-premise | High | 6–12 weeks | High (servers, patching) | High | Suitable when there is sensitive process data |
| Qlik Sense Enterprise in a managed data centre | Medium | 4–8 weeks | Shared (INFOS/client) | High | Best balance for industry with 50–500 employees |
| Qlik Cloud Analytics (dedicated tenant) | Medium | 3–6 weeks | Low | Very high | Suitable for groups with multiple factories |
The 3-year TCO rarely favours pure on-premise when the IT team has fewer than 3 people — which is the case for most Portuguese industrial SMEs. The managed data centre is often the point of balance: the client retains control over the data, INFOS manages the infrastructure.
The most common sizing error in Portugal
Factories with 80 to 150 employees buy licences for 50 Qlik users "because that's what fits the budget" and then discover that real self-service requires the shift foreman, the quality manager and the financial controller to have simultaneous access — which exceeds the licensing. The result: the tool ends up on two people's computers and Excel returns. Size by the number of real operational decision-makers, not by the number of "power users" that IT imagines at proposal time.
There is a second, less visible error: buying Analyzer (read) licences for users who in practice need Creator (dashboard editing). The distinction seems administrative. In practice, a shift foreman who cannot adjust the date filter on their own dashboard will stop using the tool within three weeks.
Decision matrix: when to move to self-service and when not to
| Criterion | Self-service viable | Self-service premature |
|---|---|---|
| Quality of source data | ERP with consistent data for more than 12 months | Multiple systems without integration, duplicate data |
| Data literacy in the operation | At least 1 person per department who reads charts | All data reading goes through IT or management |
| Governance defined | Data owners identified by area | No clear person responsible for the data |
| Volume of active users | More than 10 regular users expected | Fewer than 5 users — fixed dashboards are sufficient |
| Decision frequency | Daily or weekly operational decisions | Monthly reports for management — PDF is enough |
| Integration with ERP | Direct connector or structured ETL | Manual Excel exports as the main source |
What works in practice: three sector patterns
Pattern 1 — Textiles: OEE per shift without going through IT
In a typical knitwear factory in the Guimarães cluster, with production across three shifts and foremen who change at 6am, 2pm and 10pm, the pattern that works is as follows: KORA Productivity captures stoppage and production data in real time; Qlik Sense publishes a shift dashboard that the foreman accesses on an industrial tablet before signing the incident log. No ticket to IT. No email with Excel. The next shift's foreman sees the exact status on arrival.
What makes this work is not the software. It is the decision that the foreman has the authority to read and act on the data, without intermediate validation. Without that organisational decision, the tablet stays in the drawer — and we have seen this happen in more than one implementation where management installed the tool but did not delegate the authority to use it.
Pattern 2 — Footwear: collection analysis without the sales team's Excel
In the footwear sector in Felgueiras, international buyers visit twice a year. The week before the visit is chaotic: the sales team wants to know which models have the highest margin, which ones have sample stock, which ones had complaints in the previous collection. In a well-made self-service configuration, the sales team carries out that analysis directly in Qlik — with filters by collection, by market, by last — without depending on the controller to prepare an ad hoc report.
The prerequisite is that the ERP MULTI has the three SKU axes correctly modelled and integrated into the Qlik dataset. When that is not done, self-service produces incomplete analyses that the sales team rejects on first use — and Excel returns. Modelling the product axes in the ERP is not a technical detail: it is the entry condition for footwear BI to work.
Pattern 3 — Distribution: ABC rotation analysis without the monthly report
In a distribution warehouse in the Lousada/Paços de Ferreira corridor, with 15,000 references and daily picking, the warehouse manager has no time to wait for the monthly rotation report. The pattern that works: a Qlik dashboard with ABC analysis updated daily, integrated with the KORA Inventory Suite, accessible on a tablet in the warehouse. The manager decides product relocations based on 24-hour data, not 30-day data.
The detail the manuals do not mention: the dashboard has to be available offline or in low-latency mode. A warehouse with patchy Wi-Fi coverage that depends on a cloud dashboard with a 3-second load per interaction will abandon the tool within two weeks. Before any Qlik configuration in a warehouse, audit network coverage point by point — including the cold storage and loading dock areas, where the signal tends to drop.
How to implement self-service BI in an industrial context: the real sequence
- Audit the data sources before installing Qlik. Identify the three to five metrics that the operation consults weekly and trace them back to the source in the ERP. If you cannot carry out this tracing, self-service will amplify the confusion, not resolve it.
- Define the data owners by operational area. Production, warehouse, sales and finance each need a person responsible who validates whether the figures make sense — not IT. This definition has to happen before training, not after.
- Configure Row-Level Security from day one. Do not leave it for "later". In a GDPR context, individual productivity data without access control is an immediate risk. Qlik allows RLS to be defined per data section — use it in the initial configuration, not as a retrofit.
- Launch with three dashboards, not thirty. The most common mistake is to build a complete catalogue at the outset. Launch the OEE-per-shift dashboard, the overdue-orders one and the margins-by-product one. Let users ask for the rest — what is not requested within 90 days is probably not needed.
- Measure adoption, not satisfaction. The success indicator is not "users liked the training". It is the number of active sessions per week per licensed user. Below two weekly sessions per user, self-service is not working — regardless of what is said in the follow-up meetings.
Post-implementation success metrics
What to measure in the first 90 days
Self-service BI adoption has a predictable mortality pattern: a spike of enthusiasm in the first two weeks, a sharp drop in the fourth week when users encounter the first piece of data that "doesn't add up", and stabilisation — upwards or downwards — between the sixth and twelfth week. What determines the direction of the stabilisation is how quickly the wrong-data problem is resolved. Configure automatic alerts in Qlik for data anomalies — unexpected null values, time-series breaks — and define an internal SLA of 48 hours for resolution. Without this mechanism, every piece of wrong data is an argument to go back to Excel.
Adoption KPIs worth monitoring
- Active sessions per licensed user per week (target: three or more)
- Percentage of ad hoc reports requested from IT that exist as self-service dashboards (target: above 70% within 6 months)
- Average time between the business question and the answer (before/after — qualitative if you have no baseline)
- Number of dashboards created by non-IT users (an indicator of real autonomy)
To explore the link between BI and operational goals in more depth, the article Industrial KPIs in Qlik Sense: what to measure and what to ignore details which metrics actually change behaviour on the factory floor — and which are noise. To understand how to structure the objectives that feed these dashboards, see OKR in production: linking factory goals to the dashboard without bureaucracy.
The sequence almost everyone reverses
In 2025, only 53.7% of companies in Portugal used an ERP (INE, 2025). Almost half of Portuguese industrial companies do not have an integrated data core — and are considering self-service BI over scattered sources. In that context, self-service does not solve the problem: it defers it with a visualisation layer over data that remains inconsistent.
The correct sequence is always the same: integrated ERP first, data quality second, self-service BI third. Whoever reverses this order buys a data analysis tool that has no data to analyse. We see this happen often enough for it to be a pattern, not an exception.
Qlik Sense is an amplification tool. It amplifies good decisions when the data is good. It amplifies confusion when it is not.
For those assessing the funding of the implementation, the article PT2030 and industrial ERP: what is eligible and how to apply covers what the PT2030/COMPETE instruments cover — and what they do not — for this type of project. To understand how BI articulates with the predictive analytics layer that is starting to appear in more advanced Portuguese factories, the article Industrial BI with Qlik Sense: from raw data to operational decision develops the complete architecture.
Self-service on the factory floor is not a software question. It is a question of who has the authority to read the data, who is responsible for its quality, and whether the organisation is willing to decentralise operational intelligence. Qlik Sense solves the technical part in weeks. The organisational part takes as long as it takes management to decide that it is worthwhile — and that time is the only time that is not in the project schedule.
Sources
- INE — Survey on the Use of Information and Communication Technologies in Enterprises, 2025. Available at: ine.pt
- Directive (EU) 2022/2555 of the European Parliament and of the Council (NIS2), transposed into Portuguese law by Decree-Law 65/2025
- Regulation (EU) 2016/679 (GDPR) and Law 58/2019 — national implementation of the GDPR
- AIMMAP / Metal Portugal — sector data for the Portuguese metallurgical and metalworking industry. Available at: aimmap.pt
Frequently asked questions
Does Qlik Sense solve scattered data problems in a factory on its own?
No. Qlik Sense solves the technical part of aggregation and visualisation, but it does not eliminate the inconsistency of the source data. If the data comes from scattered sources without prior cleaning, the software amplifies the problem instead of solving it. Data governance and an operational Data Steward are required before any implementation.
What is the greatest obstacle to self-service BI in Portuguese industry?
It is not technological, it is organisational. It has three faces: a poorly configured permissions architecture, the overestimated quality of the data source, and the lack of a decision about who "owns" each dashboard. Most companies defer this definition to after installation and never make it.
What does "self-service without governance" mean and why does it fail?
Self-service does not mean the absence of governance — it means decentralised governance. When this distinction is not spelled out at the outset, the project fails. Typically, IT installs the tool, trains two people, and six months later there are dozens of outdated dashboards that no one trusts.
Who should be the "data owner" in a Portuguese industrial SME?
There should be three distinct roles: the operational Data Steward (a foreman who validates figures on the ground), the Data Engineer (IT with a limited scope over the ERP-Qlik flows), and the Dashboard Owner (a business user who publishes views without programming). Without this separation, self-service collapses into centralised dependency.
How does the Qlik Sense associative engine differ from other tools?
It allows the user to navigate across dimensions in real time without predefined queries. A controller can filter overdue orders by customer and week, associating raw material costs, without depending on IT. This is operationally critical in industries with multiple variables, such as footwear or textiles.
What compliance risks exist with individual productivity dashboards?
If the dashboards include productivity data per operator, attendance or breaks, this is personal data under the GDPR and Law 58/2019. Qlik has Row-Level Security, but it is not configured by default. Without explicit RLS, there is regulatory non-compliance, regardless of whether the software technically allows it.
Does Qlik Sense connect to industrial machines directly?
It connects to multiple sources: ERP, CSV, SQL databases, REST APIs. But for machine data via OPC-UA, the ETL work must be done upstream, usually by tools such as KORA Productivity that already structure the data. Qlik does not turn dirty data into clean insights — it amplifies what exists.