September 29, 2026
Your Data Needs a User Manual: Semantic Layers & Headless BI
Presenters
The webinar starts from the symptoms of data without a user manual: shadow semantics, metric drift and Power BI models that break every week. It frames semantics as data product metadata that lives in version control next to the pipeline code and is exposed through a serving layer. Four patterns seen in practice are compared: in-tool semantics, adapted semantics, platform-unified semantics (Databricks, Snowflake, Fabric) and a composable semantic engine with open definitions served through Cube. These are mapped to three company profiles. The recommendation has three parts: shift semantic definitions left, decide how headless the organisation should be (curated core metrics, free last-mile analysis), and let LLMs draft definitions that humans approve through pull requests. Two demos and a live Q&A with audience polls close the session.
