https://linen.dev logo
Join Slack
Powered by
[ volume anomaly with dimension breakdown reportin...
# community-support
o
[ volume anomaly with dimension breakdown reporting discrepancy ) Context: • We introduced a model called 
time_series_mnt_partner
that aggregates the revenue over time broken down by partner_id for over 100 partners. • Building on the time_series_mnt_partner model, we implemented a volume anomaly detection test with dimension breakdown. Problem: • The model flagged an anomaly with a single partner yesterday, but the result description and the values in the error are not comparable, and we cannot interpret them.
the results description
Copy code
config:
      elementary:
        timestamp_column: date
    columns:
      - name: revenue
        tests:
          - elementary.column_anomalies:
              name: partner_revenue_anomalies
              meta:
                description: "Daily interaction count anomaly test with partner revenue breakdown"
              column_anomalies:
                - sum
              dimensions:
                - partner_id
              time_bucket:
                period: day
                count: 1
              anomaly_sensitivity: 2
              detection_period:
                period: week
                count: 1
              training_period:
                period: week
                count: 4
              detection_delay:
                period: week
                count: 1
              seasonality: day_of_week
cc @Dmitrii Khriashchikov
10 Views
Open in Slack
PreviousNext
Powered by