Why Store test failures for analysis in dbt? - Purpose & Use Cases

The Big Idea

What if you could turn messy test failures into clear insights with just a simple table?

The Scenario

Imagine running many data quality tests manually and writing down each failure in a notebook or spreadsheet.

Later, you want to find patterns or fix recurring issues, but the notes are scattered and inconsistent.

The Problem

Manually tracking test failures is slow and error-prone.

You might miss important details or lose track of when and why a test failed.

This makes it hard to analyze trends or improve data quality over time.

The Solution

Storing test failures automatically in a structured table lets you easily query and analyze them.

You can track failure history, identify common problems, and prioritize fixes efficiently.

Before vs After

✗ Before

Run tests -> Copy failures to spreadsheet -> Search manually

✓ After

Create failure table -> Insert failures automatically -> Query failures with SQL

What It Enables

You can quickly spot data issues, track improvements, and make data quality a continuous, measurable process.

Real Life Example

A data team stores all test failures in a table and finds that a specific source system causes 70% of errors, so they focus their efforts there.

Key Takeaways

Manual tracking is slow and unreliable.

Automated storage of failures enables easy analysis.

Improves data quality by making issues visible and actionable.

Practice

1. What does setting store_failures = true in a dbt test configuration do?

easy

A. It sends an email notification when a test fails.

B. It skips the test execution entirely.

C. It deletes the rows that fail the test from the source table.

D. It saves the rows that fail the test for later analysis.

Solution

1. Step 1: Understand the purpose of store_failures

   This setting controls whether dbt saves the rows that fail a test.

2. Step 2: Identify the effect of setting it to true

   When set to true, dbt stores failing rows in a table for analysis.

3. Final Answer:

   It saves the rows that fail the test for later analysis. -> Option D

4. Quick Check:

   store_failures = true means save failing rows [OK]

Hint: Remember: store_failures saves failing rows for review [OK]

Common Mistakes:

• Thinking it skips tests instead of saving failures
• Assuming it deletes data on failure
• Confusing it with notification settings

2. Which of the following is the correct way to enable storing failures in a dbt test config?

easy

A. config(store_failures == true)

B. config(store_failures: true)

C. config(store_failures = true)

D. config(store_failures = "true")

Solution

1. Step 1: Recall dbt config syntax

   dbt uses Python-like syntax with keyword arguments inside config().

2. Step 2: Identify the correct assignment

   The correct syntax uses equals sign and boolean true without quotes.

3. Final Answer:

   config(store_failures = true) -> Option C

4. Quick Check:

   Use equals and boolean true without quotes [OK]

Hint: Use equals sign and true without quotes in config() [OK]

Common Mistakes:

• Using colon instead of equals sign
• Using double equals (==) instead of single equals
• Putting true in quotes making it a string

3. After running a dbt test with store_failures = true, which SQL query will show the failing rows for a model named orders and test unique?

medium

A. SELECT * FROM dbt_test__orders__unique

B. SELECT * FROM orders WHERE test = 'unique' AND status = 'fail'

C. SELECT * FROM dbt_failures.orders_unique

D. SELECT * FROM test_failures.orders_unique

Solution

1. Step 1: Recall naming convention for failure tables

   dbt stores failures in tables named dbt_test__model__test.

2. Step 2: Apply model and test names

   For model 'orders' and test 'unique', the table is dbt_test__orders__unique.

3. Final Answer:

   SELECT * FROM dbt_test__orders__unique -> Option A

4. Quick Check:

   Failure table = dbt_test__model__test [OK]

Hint: Use dbt_test__model__test naming to find failure tables [OK]

Common Mistakes:

• Assuming failure data is in original model table
• Using incorrect table prefixes
• Trying to filter failures in source tables

4. You set store_failures = true in your test config, but after running tests, no failure tables appear. What is the most likely cause?

medium

A. The test did not find any failing rows, so no failure table was created.

B. You must manually create failure tables before running tests.

C. The store_failures setting only works for schema tests, not data tests.

D. Failure tables are stored in a different schema and need to be queried there.

Solution

1. Step 1: Understand when failure tables are created

   Failure tables are only created if the test finds failing rows.

2. Step 2: Check if test found failures

   If no failures, no failure table is generated.

3. Final Answer:

   The test did not find any failing rows, so no failure table was created. -> Option A

4. Quick Check:

   No failures means no failure table [OK]

Hint: No failure rows means no failure table created [OK]

Common Mistakes:

• Thinking failure tables must be pre-created
• Assuming store_failures only works for some test types
• Looking in wrong schema without verifying failures

5. You want to analyze failing rows from multiple tests on the customers model. Which approach best helps you combine all failure data for analysis?

hard

A. Manually export failure tables and merge them outside the database.

B. Create a union query combining all dbt_test__customers__* failure tables.

C. Run each test again with store_failures = false to avoid duplicates.

D. Use the original customers table and filter rows with errors.

Solution

1. Step 1: Identify failure tables for each test

   Each test creates a failure table named dbt_test__customers__testname.

2. Step 2: Combine failure tables using SQL UNION

   Use a UNION query to combine all failure rows from these tables into one result set.

3. Final Answer:

   Create a union query combining all dbt_test__customers__* failure tables. -> Option B

4. Quick Check:

   Union failure tables to analyze all failures together [OK]

Hint: Use UNION on all failure tables for combined analysis [OK]

Common Mistakes:

• Disabling store_failures loses failure data
• Merging outside database is inefficient
• Filtering original table misses failure details