Semantična podatkovna plast: manjkajoči temelj analitike nove generacije

Organizations no longer have just too much data, but often too many dashboards and reports, among which it is difficult to recognize what really brings value or points out a problem. The paper discusses the semantic data layer as a common semantic basis above verified data: a managed layer of metrics, dimensions, concepts, relationships and rules that can be consistently used by BI tools, applications and artificial intelligence systems. We show the development from data warehouse and analytical models to reusable semantics and explain the difference between star schema, UML model, ontology and knowledge graph. The reference architecture is illustrated with the example of sales margin analysis. On this basis, we present a data assistant that reviews multiple analytical views, extracts important signals and suggests focus, but does not make decisions on its own. The paper presents a practical architectural framework and not an empirical validation of a specific product.