Why Test severity levels in dbt? - Purpose & Use Cases

The Big Idea

What if you could instantly know which data problems need your attention right now?

The Scenario

Imagine you run a data project where you check your data quality by manually scanning error logs and spreadsheets to find problems.

You get many alerts but can't tell which ones are urgent or which can wait.

The Problem

This manual way is slow and confusing.

You waste time fixing small issues that don't matter much, while missing big problems that break your reports.

Errors pile up and you lose trust in your data.

The Solution

Test severity levels let you label data tests as error, warn, or info.

This helps you focus on fixing the most important problems first.

It organizes your alerts so you know what to act on immediately and what to monitor.

Before vs After

✗ Before

run all tests and check logs manually

✓ After

severity: error\nseverity: warn\nseverity: info

What It Enables

With test severity levels, you can quickly spot and fix the biggest data issues before they cause harm.

Real Life Example

A data analyst uses severity levels to catch critical data mismatches that would break dashboards, while ignoring minor warnings that don't affect decisions.

Key Takeaways

Manual checking mixes all problems together, causing confusion.

Severity levels prioritize data tests by importance.

This saves time and improves data trust.

Practice

1. What does setting a dbt test severity level to ERROR do?

easy

A. It stops the dbt run if the test fails.

B. It only logs a warning but continues the run.

C. It ignores the test result completely.

D. It retries the test automatically.

Solution

1. Step 1: Understand dbt test severity levels

   dbt uses severity levels to decide what happens when a test fails.

2. Step 2: Identify the effect of ERROR severity

   When severity is set to ERROR, dbt stops the run immediately on failure.

3. Final Answer:

   It stops the dbt run if the test fails. -> Option A

4. Quick Check:

   ERROR severity = stops run [OK]

Hint: ERROR severity stops the run; WARN just warns [OK]

Common Mistakes:

• Confusing ERROR with WARN severity
• Thinking ERROR retries the test
• Assuming ERROR ignores failures

2. Which of the following is the correct way to set a test severity to WARN in a dbt YAML test configuration?

easy

A. severity: warn

B. severity = WARN

C. severity: WARN

D. severity: Warning

Solution

1. Step 1: Recall YAML syntax for dbt test severity

   dbt expects severity as a key-value pair with uppercase values like WARN or ERROR.

2. Step 2: Identify correct syntax

   The correct syntax uses a colon and uppercase WARN: severity: WARN.

3. Final Answer:

   severity: WARN -> Option C

4. Quick Check:

   YAML key-value with uppercase WARN = correct [OK]

Hint: Use colon and uppercase WARN for severity [OK]

Common Mistakes:

• Using lowercase 'warn' instead of uppercase
• Using equals sign instead of colon
• Spelling severity value incorrectly

3. Given this dbt test configuration snippet:

tests:
  - unique:
      column_name: id
      severity: WARN

What happens if the test fails during a dbt run?

medium

A. The failure is logged as a warning, but the run continues.

B. The test automatically retries until it passes.

C. The test is ignored and no message is shown.

D. The run stops immediately with an error.

Solution

1. Step 1: Analyze the severity level in the test config

   The severity is set to WARN, which means dbt should warn but not stop.

2. Step 2: Understand dbt behavior on WARN severity

   When a test fails with WARN severity, dbt logs a warning and continues the run.

3. Final Answer:

   The failure is logged as a warning, but the run continues. -> Option A

4. Quick Check:

   WARN severity = warn and continue [OK]

Hint: WARN severity warns but lets run continue [OK]

Common Mistakes:

• Thinking WARN stops the run
• Assuming WARN ignores failures
• Believing tests retry automatically

4. You wrote this test in your dbt model YAML:

tests:
  - not_null:
      column_name: user_id
      severity: ERROR

But your dbt run does not stop when the test fails. What is the likely problem?

medium

A. The test name 'not_null' is invalid.

B. The severity key is misplaced; it should be under 'config'.

C. The severity value should be lowercase 'error'.

D. The test is not actually failing.

Solution

1. Step 1: Check correct placement of severity in dbt tests

   Severity must be set inside the config block, not directly under the test.

2. Step 2: Identify why run does not stop

   Since severity is misplaced, dbt ignores it and uses default behavior, so run does not stop.

3. Final Answer:

   The severity key is misplaced; it should be under 'config'. -> Option B

4. Quick Check:

   Severity must be inside config = The severity key is misplaced; it should be under 'config'. [OK]

Hint: Put severity inside config block to take effect [OK]

Common Mistakes:

• Using lowercase severity value
• Misplacing severity outside config
• Assuming test name is wrong

5. You want to run a dbt test that warns on missing emails but errors on duplicate emails. Which YAML configuration correctly sets different severities for these tests on the email column?

hard

A. tests:
- not_null:
column_name: email
severity: WARN
- unique:
column_name: email
severity: ERROR

B. tests:
- not_null:
column_name: email
config:
severity: ERROR
- unique:
column_name: email
config:
severity: WARN

C. tests:
- not_null:
column_name: email
severity: ERROR
- unique:
column_name: email
severity: WARN

D. tests:
- not_null:
column_name: email
config:
severity: WARN
- unique:
column_name: email
config:
severity: ERROR

Solution

1. Step 1: Recall correct severity placement in dbt YAML

   Severity must be inside a config block under each test.

2. Step 2: Match severities to tests as required

   Set not_null severity to WARN and unique severity to ERROR inside their config blocks.

3. Final Answer:

   Tests with severity inside config: not_null WARN, unique ERROR. -> Option D

4. Quick Check:

   Severity inside config with correct WARN/ERROR = tests: - not_null: column_name: email config: severity: WARN - unique: column_name: email config: severity: ERROR [OK]

Hint: Use config block for different severities per test [OK]

Common Mistakes:

• Placing severity outside config block
• Swapping WARN and ERROR severities
• Using lowercase severity values