02. Working with Seeds
What Are Seeds in dbt?
Seeds are CSV files that you include in your dbt project and load into your data warehouse as tables.
They are often used for:
- Reference data (e.g., country codes, currency mappings)
- Lookup tables
- Static configurations
Why Use Seeds?
- Version-controlled static data
- No need for external ETL for small reference tables
- Easy to manage in Git and share with your team
How to Create a Seed?
- Create a folder named
data/in your dbt project. - Add a CSV file (e.g.,
currencies.csv):
currency_code,currency_name
USD,US Dollar
EUR,Euro
INR,Indian Rupee
How to Load Seeds into Your Warehouse?
Run:
dbt seed
This creates a table in your warehouse:
analytics.data.currencies
Configuring Seeds in dbt_project.yml
You can configure seeds globally:
seeds:
+schema: reference
+quote_columns: false
Or per seed file:
seeds:
my_project:
currencies:
+schema: lookup
+column_types:
currency_code: string
Applying Tests on Seeds
Define tests in schema.yml:
version: 2
models:
- name: currencies
columns:
- name: currency_code
tests:
- not_null
- unique
Best Practices for Seeds
Use seeds only for small, static datasets
Store seeds in data/ folder under version control
Keep column names lowercase and consistent
Document and test seeds like other models
Seed Use Cases
Seeds are ideal for supporting lookups, mappings, and small static datasets that are easier to maintain inside dbt than in separate ETL pipelines.
1️⃣ Country & Currency Reference
- Example: Mapping ISO country codes or currency codes to names.
country_code,country_name
US,United States
IN,India
UK,United Kingdom
2️⃣ Mapping Tables for Business Logic
- For categorizing data and applying consistent labels across models.
product_id,category
101,Electronics
102,Clothing
103,Books
3️⃣ Static Configuration Tables
- Example: thresholds, weights, or flags used in models.
metric,threshold
sales_target,50000
profit_margin,0.20
4️⃣ Hierarchy or Classification Data
- Store organization structures or product hierarchies.
dept_id,dept_name,parent_dept
10,Finance,NULL
20,Accounting,10
30,Payroll,10
5️⃣ Data Quality Rules
- Define validation rules as seeds for dynamic testing.
column_name,rule
customer_id,not_null
order_amount,positive
When NOT to Use Seeds
Large datasets (use external tables instead) Frequently changing data (use staging tables or ETL pipelines)
Summary
Seeds simplify managing static reference data within dbt. They are lightweight, version-controlled, and easily deployed.