How it Works

The all_columns_anomalies test from the Elementary dbt package applies column-level anomaly detection across every column in a model simultaneously. Rather than configuring individual column tests, this model-level test automatically monitors all columns for statistical deviations from their historical norms — covering metrics such as null rates, distinct value counts, and numeric distributions.

This test is particularly powerful for wide tables where manually testing every column would be impractical.

Steps and Conditions

1. Model Selection: Apply the test at the model level.

2. Automatic Column Coverage: All columns in the model are monitored.

3. Historical Baseline: Elementary uses a configurable training period to establish normal baselines.

4. Anomaly Detection: Statistical deviations from baseline are flagged.

5. Outcome: Pass if no column metrics deviate beyond acceptable thresholds; fail if anomalies are detected.

Example Usage: Marketing Analytics

A marketing analytics team has a wide campaign_performance table with dozens of columns. Instead of manually writing tests for each column, they apply all_columns_anomalies to monitor everything automatically.

models:
  - name: campaign_performance
    tests:
      - elementary.all_columns_anomalies:
          config:
            severity

models:
  - name: campaign_performance
    tests:
      - elementary.all_columns_anomalies:
          config:
            severity

models:
  - name: campaign_performance
    tests:
      - elementary.all_columns_anomalies:
          config:
            severity

This provides broad-spectrum data quality monitoring without requiring individual column configuration.