Author: Patrik Götz, Netzeffekt

At first glance, things look good in many communications departments: Dashboards have been built, KPIs defined, data warehouses connected. The CommTech Index shows progress, at least technically. And yet conversations on Monday mornings surprisingly often revolve around the same question: can we actually trust these figures?

Are they calculated correctly? Are these the official figures for the Board of Management? Do we need to recalculate them “just to be on the safe side” – perhaps even in Excel? At the latest when marketing and communication present the same key figure and deliver two different results, it becomes clear that the problem is no longer a lack of data. It is the lack of meaning.

When every department has its own truth

What can be observed in many places today is a creeping loss of confidence in figures. Not because they are missing, but because they remain in need of explanation. KPIs are discussed, questioned and reinterpreted. The result is a phenomenon that many people are aware of, but rarely name openly: Shadow Analytics. Departments build their own calculations, their own Excel truths – parallel to official dashboards.

The core of the problem lies deeper. Numbers do not speak for themselves. They need context. And this is where semantic models come into play – a topic that is surprisingly rarely discussed, even though it forms the foundation of data-driven organizations.

Semantic models: more than just IT documentation

The term “semantic models” originally comes from IT. There, it describes how data is technically processed, linked and made available. But this perspective falls short. Semantic models are much more than technical metadata – they are the place where meaning is created.

Four levels are decisive here:

• Technical metadata ensures correct processing: data fields, data types, links, update frequencies.
• Technical metadata creates uniformity: KPI definitions, impact levels, target values, benchmarks.
• Descriptive metadata makes data usable: comprehensible designations, meaningful dimensions, translations, thematic classifications.
• Operational metadata ensures sustainability: responsibilities, maintenance processes, validity.

Only when these levels interact does operational data become truly connectable – within the organization and beyond.

Why semantic models are almost always missing

If semantic models are so central, why do they exist so rarely? The reasons are structural:

They are invisible. Nobody applauds cleanly documented semantics.
They lie between responsibilities. IT, BI and business departments feel equally – and yet no one really – responsible.
Dashboards have long concealed the problem. Explanatory texts in small print are no substitute for structured meaning.
Project and time pressure replace documentation. What is seen is what delivers results quickly.
And last but not least: semantics is seen as overhead. As tedious hard work without glamor.

The result: data products scale technically – but not in terms of their comprehensibility.

Why ignoring is no longer an option

For a long time, this deficit could be covered up. But four developments make this impossible today.

Firstly, the growing maturity of data in organizations. More and more people are accessing data – and need guidance.
Secondly: The scaling of data products. The semantic proliferation increases with every new tool.
Thirdly: Large Language Models as a data access point. LLMs provide answers – but only as good as the context they are given. Without semantics, they confidently produce false truths.
Fourth: Self-service BI. If you decentralize analytical competence, you have to secure meaning centrally.

Without semantic models, data-driven work becomes a game of chance.

The chain reaction of missing meaning

If the semantics are missing, a creeping KPI drift begins. Key figures lose their comparability. Discussions replace analysis. Shadow analytics is on the rise. Self-service fails due to a lack of trust. And AI applications deliver results that no one can seriously justify.

This is not a technical problem. It is an organizational and management problem.

Semantics is communication work

A semantic model is neither a single tool nor a PowerPoint in the basement. It is structured communication work. It forces organizations to come to an understanding, take responsibility and establish common definitions – machine-readable, connectable and reusable.

Or to put it another way: when people argue about numbers, there is no lack of analysis. There is a lack of semantics.

About the author:

Partik Götz is Senior Consultant for Digital Analytics & BI at netzeffekt and has more than six years of experience in analytics engineering. He has extensive expertise in data analytics and works closely with organizations to develop data-driven strategies that meet the demands of the digital age. His core competencies include strategic consulting to measure communication effectiveness, developing and implementing data products, and creating training programs that teach professionals how to use analytical software solutions.