New message text here

You can do this with an

elementary.volume_anomalies

test configured on daily buckets, with a 1‑day detection period and

ignore_small_changes

set to 50% drop. This will effectively look at “yesterday” (last full 24h bucket) and alert if the row count drops by ≥50% vs. history. Example `schema.yml`:

models:
  - name: customers
    config:
      elementary:
        timestamp_column: created_at  # when the customer was created
    data_tests:
      - elementary.volume_anomalies:
          arguments:
            # Count new customers per day
            time_bucket:
              period: day
              count: 1

            # Detect anomalies only in the most recent full day (yesterday)
            detection_period:
              period: day
              count: 1

            # Use recent history as training window (e.g. last 14 days)
            training_period:
              period: day
              count: 14

            # Only care about drops, and only if >= 50%
            anomaly_direction: drop
            ignore_small_changes:
              drop_failure_percent_threshold: 50

Key points from the docs: •

volume_anomalies

monitors row count per time bucket; with

time_bucket: {period: day, count: 1}

it counts new rows per day. [Volume anomalies; time_bucket] • The test “only runs on completed time buckets”, so with daily buckets, if you run it during today it will evaluate yesterday’s bucket (i.e., the last full 24h). [Volume anomalies] •

ignore_small_changes.drop_failure_percent_threshold: 50

ensures the test only fails if the drop is at least 50% compared to the expected value from the training period. [ignore_small_changes example] If you also have strong day‑of‑week patterns, you can add:

seasonality: day_of_week

so each day is compared to previous same weekdays. [Seasonality usage] 👍 0 button 👎 0 button ℹ️ Tip: Mention @elementary.ai in the thread for follow-ups or DM the app for 1-to-1 help.

Elementary’s governance-related AI recommendations for tests come from two main agents in Elementary Cloud, and you review them directly in the UI rather than downloading a separate rules file. 1. Governance AI Agent (governance-focused, including metadata/test gaps) ◦ Open the Catalog in Elementary Cloud. ◦ Trigger the Governance Agent (e.g., ask “Which tables in the marketing domain are missing ownership?”). ◦ The agent analyzes metadata, tags, ownership, and conventions, then surfaces actionable suggestions (including gaps that often imply missing tests). ◦ When you accept suggestions, it can open a pull request with the proposed metadata changes so you can review them in your repo. [Governance agent] 2. Test Recommendation AI Agent (direct test implementation guidance) ◦ From the test configuration screen (or the Test Coverage view), open the Test Recommendation Agent. [Test coverage; Test recommendation agent] ◦ Ask questions like “What tests should I add for this model?” or “Which upstream tables need tests to protect this dashboard?”. ◦ The agent analyzes lineage, metadata, SQL logic, and existing tests, then shows recommended tests in the UI. ◦ Once you approve, it automatically opens a PR with the test configurations (YAML/dbt tests) so you can review and merge them. [Test recommendation agent; Config as code] All AI agents (Ella) work this way: recommendations are visible in the Cloud UI, and any implementation is done via pull requests to your connected repo—there’s no separate downloadable “governance ruleset” described in the docs. [Ella overview; Config as code] 👍 0 button 👎 0 button ℹ️ Tip: Mention @elementary.ai in the thread for follow-ups or DM the app for 1-to-1 help.