Testing Data Quality

Ensuring data quality is critical for any analytics pipeline. dbt provides built-in testing capabilities that help catch issues early, enforce data integrity, and maintain confidence in your transformations. This guide explains how to implement tests in your dbt project.

Why Testing Matters in dbt

Data tests serve several essential purposes:

• Validate assumptions about your data

• Catch errors before they impact downstream consumers

• Document expectations about data properties

• Ensure consistency across transformations

Without testing, issues can creep into your data pipeline, potentially leading to incorrect business decisions or loss of trust in your analytics.

Benefits of dbt Testing

• Ensures Data Integrity – Prevents duplicates, null values, and referential mismatches.

• Validates Business Logic – Confirms that data meets expected criteria.

• Catches Issues Early – Detects errors before they affect downstream analytics.

• Automates Quality Checks – Reduces the need for manual data validation.

• Supports Collaboration – Helps teams align on data expectations.

---

Types of dbt Tests

dbt supports two main types of tests:

1. Generic Tests (Built-in)

Generic tests are reusable test definitions that can be applied to multiple models and columns. dbt includes four built-in generic tests:

Test	Purpose	Example Use
unique	Ensures a column has no duplicate values	Primary keys, email addresses
not_null	Ensures a column contains no NULL values	Required fields, join keys
accepted_values	Validates that column values are within a specified list	Status fields, categories
relationships	Ensures referential integrity between tables	Foreign keys, dimensional references

2. Singular Tests

Singular tests are custom SQL queries that define specific test logic. These are one-off tests written as SQL queries that return failing records.

---

Adding Tests to Your Models

Tests in dbt are typically defined in YAML files alongside your models.

Generic Tests Example

# models/schema.yml
version: 2

models:
  - name: customers
    columns:
      - name: customer_id
        tests:
          - unique
          - not_null
      - name: email
        tests:
          - unique
          - not_null
      - name: status
        tests:
          - accepted_values:
              values: ['active', 'inactive', 'pending']
      - name: country_id
        tests:
          - relationships:
              to: ref('countries')
              field: id

This YAML configuration:

• Tests that customer_id values are unique and not null

• Tests that email values are unique and not null

• Tests that status values are only 'active', 'inactive', or 'pending'

• Tests that each country_id exists in the countries table

Singular Test Example

Singular tests are SQL files in the tests/ directory:

-- tests/assert_total_payment_amount_matches_order_amount.sql
-- This test checks that payment amounts sum to order amounts
SELECT
  orders.order_id,
  orders.amount as order_amount,
  SUM(payments.amount) as payment_amount
FROM {{ ref('orders') }}
LEFT JOIN {{ ref('payments') }} ON orders.order_id = payments.order_id
GROUP BY orders.order_id, orders.amount
HAVING ABS(orders.amount - SUM(COALESCE(payments.amount, 0))) > 0.01

This test identifies orders where the payment amounts don't match the order amount.

---

Running Tests

dbt makes it easy to run tests as part of your workflow.

Operation	Command	Description
Running All Tests	dbt test	Runs all tests in your dbt project
Testing Specific Models	dbt test --models customers	Runs all tests associated with a specific model
Running a Single Test	dbt test --select test_name	Runs a specific test by name
Testing Critical Models Only	dbt test --select tag:critical	Runs tests only for models tagged as 'critical'

---

Test Configuration Options

You can configure how tests behave using additional parameters.

Setting Severity Levels

Tests can be warnings instead of errors:

models:
  - name: orders
    columns:
      - name: status
        tests:
          - accepted_values:
              values: ['completed', 'shipped', 'returned']
              severity: warn  # Won't cause pipelines to fail

Store Test Failures

Save test failures for analysis:

models:
  - name: large_table
    columns:
      - name: id
        tests:
          - unique:
              config:
                store_failures: true  # Saves failures to a table

Limiting Failure Volume

Control how many failures are reported:

models:
  - name: events
    columns:
      - name: event_id
        tests:
          - unique:
              config:
                limit: 100  # Only show first 100 failures

---

Creating Custom Generic Tests

You can extend dbt's testing capabilities by creating custom generic tests as macros:

-- macros/test_is_valid_email.sql
{% macro test_is_valid_email(model, column_name) %}

select *
from {{ model }}
where 
    {{ column_name }} is not null 
    and not regexp_like(
        {{ column_name }}, 
        '^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}$'
    )

{% endmacro %}

Then use it just like built-in tests:

models:
  - name: customers
    columns:
      - name: email
        tests:
          - is_valid_email

---

Test Organization Strategies

As your project grows, organizing tests becomes important:

Test by Business Domain

Group tests alongside the models they validate:

models/
├── marketing/
│   ├── schema.yml      # Contains tests for marketing models
│   ├── campaigns.sql
│   └── ad_performance.sql
└── finance/
    ├── schema.yml      # Contains tests for finance models
    ├── transactions.sql
    └── accounts.sql

Centralized Tests

Maintain all tests in a dedicated location:

models/
└── ...
tests/
├── generic/            # Custom generic tests
│   └── is_valid_email.sql
└── singular/           # Singular tests
    ├── marketing/
    │   └── campaign_consistency.sql
    └── finance/
        └── transaction_reconciliation.sql

---

Troubleshooting Failed Tests

When tests fail, dbt provides information to help diagnose the issue:

1. Review test SQL: dbt outputs the SQL for the failing test

2. Examine failing records: Look at examples of failing records

3. Check compiled SQL: Review the compiled test SQL in target/compiled/

4. Store failures: Use store_failures: true to analyze failure patterns

---

Test Output Examples

Passing Tests

18:42:10 | 1 of 4 START test not_null_orders_order_id ............................ [RUN]
18:42:10 | 1 of 4 PASS not_null_orders_order_id ................................. [PASS in 0.08s]
18:42:10 | 2 of 4 START test unique_orders_order_id ............................. [RUN]
18:42:10 | 2 of 4 PASS unique_orders_order_id ................................... [PASS in 0.10s]

Failing Tests

18:42:11 | 3 of 4 START test accepted_values_orders_status_completed__shipped__returned ... [RUN]
18:42:11 | 3 of 4 FAIL accepted_values_orders_status_completed__shipped__returned ... [FAIL in 0.15s]
18:42:11 | 4 of 4 START test relationships_orders_customer_id__customer_id__ref_customers_ ... [RUN]
18:42:11 | 4 of 4 FAIL relationships_orders_customer_id__customer_id__ref_customers_ ... [FAIL in 0.19s]

For failing tests, dbt shows details about the failures:

Failure in test relationships_orders_customer_id__customer_id__ref_customers_
Got 2 results, expected 0.

compiled SQL at target/compiled/.../relationships_orders_customer_id__customer_id__ref_customers_.sql

---

Best Practices for dbt Testing

Test Coverage Strategy

• Test primary keys with unique and not_null

• Test foreign keys with relationships

• Test categorical fields with accepted_values

• Test business logic with singular tests

• Focus on testing critical data first

Test Organization

• Use a consistent naming convention for singular tests

• Group related tests together

• Document what each test validates

Test Execution

• Run tests before finalizing model changes

• Include tests in CI/CD pipelines

• Alert on test failures in production

Test Maintainability

• Prefer generic tests for common validations

• Create custom generic tests for repeated patterns

• Use macros to generate complex test logic {% endhint %}

By implementing a robust testing strategy in dbt, you can ensure your data transformations maintain high quality and reliability, building trust in your analytics data among stakeholders.