Semantic Model
A semantic model is a structured definition of business entities, their attributes, relationships, and calculations, providing a unified data structure for analytics and reporting.
A semantic model represents how business entities (customers, orders, products, accounts) relate to each other and what properties they have. It's more granular than a raw database schema but more structured than a free-form data dictionary. The semantic model specifies: what entities exist, which tables represent them, how they join, which fields are dimensions or measures, and what calculations apply.
Semantic models solve ambiguity and inconsistency in multi-table analytics. When analysts work with raw schemas, they make different assumptions about join paths, cardinality, and aggregation. A semantic model codifies these assumptions: it formally states that an order belongs to exactly one customer, that orders have items (one-to-many), and that revenue always aggregates across items before orders. This prevents double-counting and logic errors.
Semantic models are typically defined in YAML or through UI tools and are language-agnostic. They drive metric generation, query optimization, and governance enforcement. A semantic model is the blueprint that the metric layer builds upon. Tools like dbt, Looker, and Tableau embed semantic model concepts under different names (metrics definitions, explores, semantic layers), but all follow the principle of modeling business logic explicitly rather than embedding it in queries.
Key Characteristics
- ▶Defines entities, attributes, and their relationships
- ▶Specifies table mappings and join logic
- ▶Classifies fields as dimensions, measures, or keys
- ▶Documents cardinality and valid aggregations
- ▶Enables calculation of derived measures
- ▶Version-controlled and auditable
Why It Matters
- ▶Correctness: Enforced join paths eliminate double-counting errors
- ▶Consistency: Single definition of customer, order, product across tools
- ▶Reusability: Models serve multiple BI tools and use cases
- ▶Documentation: Semantic models act as executable business documentation
- ▶Automation: Models drive schema validation and metric generation
Example
A semantic model defines: customers (one-to-many) to orders, orders (one-to-many) to order items, items (many-to-one) to products. The model specifies that revenue must aggregate items before orders to avoid duplication. Any tool using this model automatically generates correct aggregations rather than relying on analyst vigilance.
Coginiti Perspective
Coginiti's semantic model is defined in SMDL (Semantic Model Definition Language), an HCL-like syntax stored in .smdl files. Each entity maps to a database table or SQL query, with dimensions and measures declared as typed attributes. Relationships between entities specify cardinality (one_to_one, one_to_many, many_to_one) and join expressions using semantic identifiers rather than physical column names. The semantic model is version-controlled through the Analytics Catalog, where changes follow a promotion workflow from personal to shared to project hub workspaces.
More in Semantic Layer & Metrics
Business Logic Layer
A business logic layer is the component of a semantic layer or data system that encodes business rules, calculations, and transformations, making them reusable and enforced across analytics.
Data Abstraction Layer
A data abstraction layer is a software or architectural component that sits between raw data sources and analytics consumers, providing unified access and hiding implementation complexity.
Data Semantics
Data semantics refers to the documented meaning, business context, and valid usage of data elements, including definitions, relationships, constraints, and governance rules.
Derived Metrics
Derived metrics are metrics calculated from other base metrics or dimensions rather than directly from raw fact tables, enabling metric composition and reducing calculation redundancy.
Dimension
A dimension is a categorical or descriptive attribute used to slice, filter, and organize metrics, such as product, region, customer segment, or date.
Governed Metrics
Governed metrics are business metrics with centrally defined calculations, owners, approval workflows, and enforced standards that ensure consistency and trustworthiness across all analytics consumers.
See Semantic Intelligence in Action
Coginiti operationalizes business meaning across your entire data estate.