Exposures and Semantic Layer

Semantic Layer Architecture

Metrics Flow

+-----------------------------------------------------------------------------+
|                     METRICS DEFINITION AND USAGE                            |
+-----------------------------------------------------------------------------+
|                                                                             |
|  SEMANTIC MODEL                                                             |
|  +---------------------------------------------------------------------+    |
|  |  name: orders_semantic                                              |    |
|  |  defaults:                                                           |   |
|  |    agg_time_dimension: order_date                                    |   |
|  |  entities:                                                           |   |
|  |    - name: order_id                                                 |    |
|  |      type: primary                                                  |    |
|  |    - name: customer_id                                              |    |
|  |      type: foreign                                                  |    |
|  |  dimensions:                                                         |   |
|  |    - name: order_date                                               |    |
|  |      type: time                                                     |    |
|  |      type_params:                                                   |    |
|  |        time_granularity: day                                         |   |
|  |  measures:                                                           |   |
|  |    - name: order_total                                              |    |
|  |      type: sum                                                      |    |
|  |      expr: amount                                                   |    |
|  +---------------------------------------------------------------------+    |
|                              |                                              |
|                              v                                              |
|  METRICS DEFINITION                                                         |
|  +---------------------------------------------------------------------+    |
|  |  - name: total_revenue                                              |    |
|  |    type: simple                                                     |    |
|  |    type_params:                                                     |    |
|  |      measure: order_total                                           |    |
|  |    description: "Total revenue from all orders"                     |    |
|  |                                                                     |    |
|  |  - name: avg_order_value                                            |    |
|  |    type: derived                                                    |    |
|  |    type_params:                                                     |    |
|  |      expr: total_revenue / order_count                              |    |
|  |    description: "Average order value"                               |    |
|  +---------------------------------------------------------------------+    |
|                              |                                              |
|                              v                                              |
|  QUERY INTERFACE                                                            |
|  +---------------------------------------------------------------------+    |
|  |  metrics: [total_revenue, order_count]                              |    |
|  |  group_by: [order_date__day, customer_segment]                      |    |
|  |  where: "order_date >= '2024-01-01'"                                |    |
|  |                                                                     |    |
|  |  Result:                                                             |   |
|  |  +------------+---------+----------+------------+                  |     |
|  |  | Date       | Segment | Revenue  | Orders     |                  |     |
|  |  +------------+---------+----------+------------+                  |     |
|  |  | 2024-01-01 | Premium | $150,000 | 1,500      |                  |     |
|  |  | 2024-01-01 | Standard| $85,000  | 2,100      |                  |     |
|  |  +------------+---------+----------+------------+                  |     |
|  +---------------------------------------------------------------------+    |
|                                                                             |
+-----------------------------------------------------------------------------+

Exposure Dependencies

Detailed Explanation

The dbt semantic layer provides a standardized way to define metrics, dimensions, and business logic that can be consumed by downstream applications.

What are Semantic Models?

Semantic models are the foundation of the semantic layer. They define:

1. Entities: Keys that identify records (primary, foreign, unique)
2. Dimensions: Qualitative attributes for grouping and filtering
3. Measures: Quantitative values that can be aggregated

What are Metrics?

Metrics are defined calculations based on semantic models:

What are Exposures?

Exposures define how data is consumed downstream:

• Dashboards: BI tool dashboards (Looker, Tableau, PowerBI)
• Applications: Data applications and APIs
• Exports: Data exports to external systems

What are the Benefits of the Semantic Layer?

1. Single source of truth: Consistent metric definitions
2. Governance: Centralized business logic management
3. Reusability: Metrics defined once, used everywhere
4. Lineage: Track metric dependencies and usage
5. Performance: Pre-computed aggregations

Key Takeaway: The semantic layer standardizes metrics and dimensions, enabling consistent business logic across all downstream applications.

Code Examples

Semantic Model Definition

# semantic/order_semantic.yml
version: 2

semantic_models:
  - name: orders_semantic
    description: "Semantic model for order analytics"
    defaults:
      agg_time_dimension: order_date
    model: ref('fct_orders')
    
    entities:
      - name: order_id
        type: primary
        expr: order_id
      
      - name: customer_id
        type: foreign
        expr: customer_id
    
    dimensions:
      - name: order_date
        type: time
        type_params:
          time_granularity: day
      
      - name: order_status
        type: categorical
        expr: status
      
      - name: customer_segment
        type: categorical
        expr: segment
    
    measures:
      - name: order_total
        description: "Total revenue from orders"
        agg: sum
        expr: amount
      
      - name: order_count
        description: "Count of orders"
        agg: count
        expr: order_id
      
      - name: avg_order_value
        description: "Average order value"
        agg: average
        expr: amount

Metrics Definition

# semantic/metrics.yml
version: 2

metrics:
  - name: total_revenue
    description: "Total revenue from all orders"
    type: simple
    type_params:
      measure: order_total
    
    filter: |
      {{ Dimension('order_status') }} != 'cancelled'
  
  - name: order_count
    description: "Total number of orders"
    type: simple
    type_params:
      measure: order_count
  
  - name: avg_order_value
    description: "Average order value"
    type: derived
    type_params:
      expr: "total_revenue / order_count"
  
  - name: revenue_growth
    description: "Month-over-month revenue growth"
    type: derived
    type_params:
      expr: "(current_month_revenue - prev_month_revenue) / prev_month_revenue"
  
  - name: conversion_rate
    description: "Conversion rate from views to purchases"
    type: conversion
    type_params:
      measure: order_count
      conversion_measure: view_count
      window: 7 days

Exposure Definition

# exposures/order_dashboard.yml
version: 2

exposures:
  - name: order_analytics_dashboard
    type: dashboard
    description: "Main dashboard for order analytics"
    url: https://looker.example.com/dashboards/order_analytics
    
    depends_on:
      - ref('fct_orders')
      - ref('dim_customers')
      - ref('dim_products')
    
    owner:
      name: Analytics Team
      email: analytics@company.com
    
    meta:
      refresh_frequency: hourly
      data_source: Looker
  
  - name: ml_feature_store
    type: ml
    description: "ML feature store for customer segmentation"
    url: https://ml.example.com/features/customer_segmentation
    
    depends_on:
      - ref('customer_features')
      - ref('order_features')
    
    owner:
      name: Data Science Team
      email: ds@company.com
  
  - name: data_export_daily
    type: application
    description: "Daily data export to external system"
    url: https://api.example.com/exports/daily
    
    depends_on:
      - ref('fct_daily_metrics')
    
    owner:
      name: Data Engineering
      email: data-eng@company.com

Semantic Model with Joins

# semantic/order_customer_semantic.yml
version: 2

semantic_models:
  - name: order_with_customer
    description: "Orders with customer dimensions"
    defaults:
      agg_time_dimension: order_date
    model: ref('fct_orders')
    
    joins:
      - name: customers
        join_type: left
        sql: "ON {{ entity('order_id') }} = {{ entity('customer_id') }}"
    
    entities:
      - name: order_id
        type: primary
        expr: order_id
    
    dimensions:
      - name: order_date
        type: time
        type_params:
          time_granularity: day
      
      - name: customer_name
        type: categorical
        expr: "customers.customer_name"
      
      - name: customer_segment
        type: categorical
        expr: "customers.segment"
    
    measures:
      - name: total_revenue
        agg: sum
        expr: amount

Querying the Semantic Layer

# Query definition for Looker
query:
  metrics:
    - total_revenue
    - order_count
    - avg_order_value
  group_by:
    - order_date__day
    - customer_segment
  where: |
    {{ Dimension('order_date') }} >= '2024-01-01'
    AND {{ Dimension('order_status') }} = 'completed'
  order_by:
    - order_date__day
  limit: 1000

Performance Metrics

Best Practices

1. Define semantic models for all analytical entities
2. Use consistent naming for metrics and dimensions
3. Document all exposures with clear descriptions
4. Define relationships between semantic models
5. Use appropriate measures for different aggregation types
6. Filter at the metric level to ensure consistency
7. Monitor exposure usage to understand data consumption
8. Version control all semantic definitions

See Also

• dbt Documentation and Lineage — Documentation generation and metadata management
• The ref() Function — Model reference resolution and dependency management
• dbt Core Architecture — Manifest, DAG, and compilation pipeline
• Data Engineering Fundamentals — Modern data stack overview
• PySpark Data Pipelines — Building data pipelines with Spark