@kapa.ai Did this work ever get done / is there a way to create an anomaly test on a custom-defined metric? https://github.com/elementary-data/elementary/issues/1255

Yes! Elementary Cloud now supports custom metrics with anomaly detection via the

elementary.collect_metrics

test. [Metrics docs] Here's how it works: • Metrics are defined in code (like dbt tests) and collected each time you run

dbt test

. • You can monitor any metric on your data content, view it in the Metrics screen, and set up anomaly detection on it. • A metric can be collected with or without an anomaly detection test configured. Example configuration

data_tests:
  - elementary.collect_metrics:
      arguments:
        timestamp_column: your_timestamp_column
        time_bucket:
          period: day
          count: 1
        dimensions: sql_expression   # optional grouping
        metrics:
          - name: my_custom_metric
            type: monitor_type       # e.g. sum, avg, etc.
            columns: [col1, col2]
        where_expression: sql_expression  # optional filter

[Metrics docs] A few notes: • This feature is available on Elementary Cloud. • To include alerts for metric anomalies in your alert rules, edit your alert rule and check the "Metric anomalies" box under "Test Categories". • If you need to monitor a custom calculated value (e.g., a ratio or match rate between tables) that isn't a simple column, the recommended approach is still to materialize it as a dbt model first, then apply the test to it. So the short answer is: yes, this has been addressed — custom metric collection with anomaly detection is now supported in Elementary Cloud. 👍 0 button 👎 0 button ℹ️ Tip: Mention @elementary.ai in the thread for follow-ups or DM the app for 1-to-1 help.

@kapa.ai This isn't truly custom, and is limited to the monitor types available in column_anomalies, right? What if I wanted to monitor a ratio like the one described in the Github issue?

metric_name: roi
   sql: sum(revenue) / sum(cost)

You raise a valid point. Based on the knowledge sources,

elementary.collect_metrics

is indeed limited to the pre-defined monitor types — it does not support arbitrary SQL expressions like

sum(revenue) / sum(cost)

as a custom metric definition. [Metrics docs] For a fully custom SQL metric like your

roi

example, the knowledge sources do not indicate that this is natively supported. The workaround consistently recommended in the community is to materialize the calculation as a column in a dbt model first, then apply anomaly detection to that column: 1. Create a dbt model that computes your ratio:-- models/daily_roi.sql select date_column, sum(revenue) / nullif(sum(cost), 0) as roi from your_table group by date_column 2. Apply an Elementary anomaly test to the

roi

columnmodels - name: daily_roi columns: - name: roi tests: - elementary.column_anomalies: timestamp_column: date_column [Slack discussion] I'm sorry, but the knowledge sources don't contain enough information to confirm whether the specific

sql:

syntax from the GitHub issue you referenced has been implemented. I'd recommend checking the GitHub issue directly or asking in the Elementary community Slack for the latest status. 👍 0 button 👎 0 button ℹ️ Tip: Mention @elementary.ai in the thread for follow-ups or DM the app for 1-to-1 help.