Detailed definition
Semantic modelling turns technical data assets into reusable business meaning. It defines how tables connect, how metrics are calculated, which synonyms users may use, which hierarchies apply, and which users can access which data.
For AI BI, semantic modelling is the foundation that lets an assistant answer questions in business language without guessing the underlying schema.
Why it matters
Most data warehouses are optimized for storage, processing, and transformation, not for everyday business language. Semantic modelling closes that gap.
It also reduces inconsistency. When metric logic is defined once and reused across tools, teams spend less time debating which number is correct.
How it works
Data teams identify core business entities, measures, relationships, and governance rules. They map those definitions to physical sources and add descriptions, synonyms, and examples that help both people and AI systems interpret questions.
Good semantic modelling is iterative. Teams start with high-value concepts, observe user questions, and refine definitions as the business changes.
Practical examples
- Defining "active customer" with time windows, exclusions, and lifecycle rules.
- Mapping "SKU", "item", and "product" to the correct product concept.
- Creating reusable measures such as net revenue, gross margin, and average order value.
- Applying row-level permissions based on region or role.
Common pitfalls
- Semantic modelling is not a one-time documentation exercise.
- Modelling every possible concept before launch can slow adoption. Start with important questions and expand.
- AI-generated model suggestions still need ownership and review from data teams.