dbt Model Writer

Expert dbt model creation and editing skill for analytics engineering.

When to Use This Skill

Activate this skill when the user mentions:

• "Create a dbt model"
• "Write a dbt SQL file"
• "Convert this SQL to dbt"
• "Add a staging/intermediate/mart model"
• "Update this dbt model"
• "Fix this dbt model"

Core Capabilities

1. Model Creation

Create new dbt models with:

• Proper config blocks (materialization, tags, schema)
• Descriptive CTE structure
• Inline documentation comments
• Source and ref() functions
• Appropriate Jinja templating

2. Model Types

Staging Models (stg_*.sql)

• Purpose: Clean and standardize raw source data
• Materialization: Usually view
• Pattern: One-to-one with source tables
• Transformations: Column renaming, type casting, basic cleaning

{{
  config(
    materialized='view',
    tags=['staging']
  )
}}

with source as (
    select * from {{ source('raw', 'orders') }}
),

renamed as (
    select
        order_id,
        customer_id,
        order_date,
        status,
        -- Standardize column names and types
        cast(total_amount as decimal(10,2)) as total_amount,
        created_at,
        updated_at
    from source
)

select * from renamed

Intermediate Models (int_*.sql)

• Purpose: Complex business logic, joins, denormalization
• Materialization: Usually ephemeral or view
• Pattern: Combine multiple staging models
• Transformations: Business calculations, complex joins

{{
  config(
    materialized='ephemeral',
    tags=['intermediate']
  )
}}

with orders as (
    select * from {{ ref('stg_orders') }}
),

customers as (
    select * from {{ ref('stg_customers') }}
),

order_items as (
    select * from {{ ref('stg_order_items') }}
),

order_totals as (
    select
        order_id,
        count(*) as item_count,
        sum(quantity) as total_quantity,
        sum(quantity * price) as gross_total
    from order_items
    group by order_id
),

joined as (
    select
        o.order_id,
        o.customer_id,
        c.customer_name,
        o.order_date,
        ot.item_count,
        ot.total_quantity,
        ot.gross_total
    from orders o
    join customers c on o.customer_id = c.customer_id
    join order_totals ot on o.order_id = ot.order_id
)

select * from joined

Mart Models (fct_*.sql or dim_*.sql)

• Purpose: Analytics-ready tables optimized for BI tools
• Materialization: Usually table or incremental
• Pattern: Final dimensional models (facts and dimensions)
• Transformations: Aggregations, final business metrics

{{
  config(
    materialized='incremental',
    unique_key='order_id',
    cluster_by=['order_date'],
    tags=['mart', 'daily']
  )
}}

with orders as (
    select * from {{ ref('int_order_details') }}
)

select
    order_id,
    customer_id,
    order_date,
    item_count,
    total_quantity,
    gross_total,
    current_timestamp() as dbt_updated_at
from orders

{% if is_incremental() %}
    -- Only process new or updated orders
    where order_date >= (select max(order_date) from {{ this }})
{% endif %}

3. Config Block Standards

Essential config parameters:

{{
  config(
    materialized='table',  -- table|view|incremental|ephemeral
    schema='analytics',    -- custom schema
    alias='my_table',      -- custom table name
    tags=['daily', 'core'], -- organizational tags
    cluster_by=['date', 'id'], -- Snowflake clustering
    unique_key='id',       -- for incremental
    on_schema_change='sync_all_columns', -- incremental schema changes
    pre_hook="ALTER TABLE {{ this }} ...",
    post_hook="GRANT SELECT ON {{ this }} ..."
  )
}}

4. Incremental Strategy Patterns

Append-Only (Insert New Records)

{% if is_incremental() %}
    where created_at > (select max(created_at) from {{ this }})
{% endif %}

Update Existing Records (Merge)

{{
  config(
    materialized='incremental',
    unique_key='id',
    incremental_strategy='merge'
  )
}}

Delete + Insert

{{
  config(
    materialized='incremental',
    unique_key='date_day',
    incremental_strategy='delete+insert'
  )
}}

5. Documentation Standards

Always create corresponding schema.yml entries:

version: 2

models:
  - name: fct_orders
    description: "Fact table for order transactions"
    columns:
      - name: order_id
        description: "Unique order identifier"
        tests:
          - unique
          - not_null
      - name: customer_id
        description: "Foreign key to dim_customers"
        tests:
          - not_null
          - relationships:
              to: ref('dim_customers')
              field: customer_id
      - name: order_date
        description: "Date order was placed"
        tests:
          - not_null
      - name: total_amount
        description: "Total order value in USD"
        tests:
          - not_null

6. Jinja Best Practices

Use ref() for model dependencies

select * from {{ ref('stg_orders') }}

Use source() for raw data

select * from {{ source('raw_db', 'orders') }}

Environment-specific logic

{% if target.name == 'prod' %}
    -- Production-specific logic
{% else %}
    -- Dev/test logic
{% endif %}

Macros for reusable logic

{{ cents_to_dollars('price_cents') }} as price_dollars

Workflow Process

1. Understand Requirements

   • Ask about model purpose and layer (staging/intermediate/mart)
   • Identify source tables or upstream models
   • Determine materialization needs
   • Clarify any business logic

2. Read Existing Context

   • Check dbt_project.yml for project structure
   • Look for similar existing models
   • Review naming conventions in use
   • Check for available macros

3. Write the Model

   • Start with config block
   • Build CTEs from source up
   • Apply transformations step-by-step
   • Add inline comments for complex logic
   • Final select statement

4. Create Documentation

   • Generate schema.yml entry
   • Add column descriptions
   • Include appropriate tests

5. Validate

   • Check for syntax errors
   • Ensure all refs/sources exist
   • Verify config parameters
   • Confirm naming conventions

Common Patterns to Apply

Deduplication

with deduped as (
    select
        *,
        row_number() over (
            partition by id 
            order by updated_at desc
        ) as row_num
    from source
    where row_num = 1
)

Type 2 SCD (Slowly Changing Dimension)

with current_records as (
    select * from {{ ref('stg_customers') }}
),

historical_records as (
    select * from {{ this }}
    where is_current = true
),

changes as (
    select
        c.*,
        h.customer_key,
        case
            when h.customer_key is null then true
            when c.customer_name != h.customer_name then true
            else false
        end as has_changed
    from current_records c
    left join historical_records h
        on c.customer_id = h.customer_id
)
-- ... SCD logic continues

Surrogate Key Generation

{{ dbt_utils.generate_surrogate_key(['order_id', 'line_number']) }} as order_line_key

Error Handling

If encountering issues:

1. Check compiled SQL in target/compiled/
2. Verify all referenced models exist
3. Validate Jinja syntax
4. Ensure source definitions exist
5. Check for circular dependencies

Quality Checklist

Before completing, verify:

• Config block with appropriate materialization
• All CTEs have descriptive names
• Complex logic has comments
• Uses ref() for models, source() for raw data
• schema.yml entry created with tests
• Column descriptions added
• Follows project naming conventions
• No hardcoded values (use vars or macros)
• Incremental logic correct (if applicable)
• Efficient SQL (appropriate JOINs and filters)