Technical risk
An unsuitable transformation can change the granularity of information.
Load-script structure
Mapping loads, joins, concatenations, group by and filters were selected according to the relationship between sources.
Back to portfolio
CASE 03 / Operational intelligence
Operational intelligence: connecting data to interpret the process
I modelled more than 80 operational tables in Qlik Sense, combining source preparation, load-script transformations, key management and refresh routines supported by Python, SQL and VBA.
03.1 / Context
Manual consolidation of scattered sources
Quality, production, maintenance, delivery and material data lived in separate sources. Engineers repeated extraction and preparation before they could compare periods or investigate relationships between events.
Investigating the problem
The problem preceded visualization: gathering materials, quality, production, maintenance and delivery data required repeated preparation. Modeling had to make the relationships between those industrial domains explicit.
Implementation constraints
Sources had different structures and relationships. Integration had to retain history and run within the available infrastructure, with automated routines supported by a dedicated computer.
FragmentationQuality, production, maintenance, delivery and materials in separate sources.
Repeated effortManual preparation reducing the time available for investigation.
Limited horizonDifficulty comparing periods and understanding relationships between events.
03.2 / Development
Source integration and associative modelling
Python, SQL and VBA prepared or connected sources. In Qlik Load Script, mapping loads, joins, concatenations, groupings and filters handled different granularities; controlled keys and retained history kept the associative model interpretable.
- Industrial sources
Quality · production · maintenance · delivery · materials
- Supporting routines
Python · VBA · SQL — extraction and integration
- Qlik Load Script
Mapping · joins · concatenation · filters · history
- Associative model
Controlled keys · relationships across 80+ tables
- Analytical use
Dimensions · measures · decisions in context
Select a stage to explore its role in the flow.
← Scroll horizontally to explore the diagram →
03.3 / Project decisions
Associations, history and refresh routines required explicit choices.
Three decisions supported model integrity and refresh reliability.
Technical risk
Ambiguous associations can connect events without the intended relationship.
Key management
Synthetic keys were identified and resolved to preserve intentional associations and avoid ambiguity.
Technical risk
Refresh routines also depend on the availability of their execution environment.
Routine orchestration
Python and VBA supported integrations; a dedicated computer sustained automated routines within the available infrastructure.
From decision to implementation
- Prepare the sources
Python, SQL and VBA supported extraction and integration. Data preparation was separated from indicator interpretation.
- Build the relationships
Mapping loads, joins, concatenation, grouping and filters organized the load. Resolving synthetic keys preserved intentional associations.
- Make analysis available
The associative model connected industrial domains, retained history and enabled more frequent updates to support operational interpretation.
80+tables connected in the associative data model
03.4 / Evidence and outcomes
The same model began supporting analysis across quality, production and maintenance.
More than 80 tables gained a shared analytical structure, with less duplicated preparation and more historical context. The published evidence supports the model's scope and operating mechanism; it does not establish a quantified gain in accuracy or financial return.
- an integrated view across operational dimensions;
- less duplicated preparation and consolidation;
- more historical context for comparison;
- better context for analysis and decisions.
Verification / interpreting the evidence
What the evidence establishes
More than 80 tables were connected in the model. This describes integration coverage; it does not, by itself, measure accuracy or financial benefit.
Engineering criterion
Interpreting an indicator requires attention to the selected associations and time period. Connecting sources is useful only when relationships preserve the industrial meaning of the data.
Limits of the public evidence
The portfolio does not include corporate samples, reconciliation test records or an accuracy measurement series. None of these outcomes is inferred from the table count.
03.5 / Synthesis
Learning
The problem showed that source integration is more than loading tables: the model depends on preserved granularity, controlled keys and retained history. The architecture of relationships became as important as the visual layer.