Skip to content

Detect

Detect watches a KPI over time and flags the points that fall outside what was expected. It checks every step of your market equation and every segment of your dimensions, then ranks where the anomalies are.

Use Detect to monitor a dashboard: did anything unusual happen yesterday, and where?

Detect adds an Anomaly module:

  • the chart of the KPI over time, with its confidence interval and the outliers highlighted above or below it;
  • the Scores table, that ranks every KPI and segment by interest and counts its outliers;
  • the comment, that describes the outliers in view and lists the other areas to look at.

See Charts below.

  1. Choose the dimension to Detect anomaly on, usually a date.
  2. Check the market equation: Detect flags each step, and the total KPI.
  3. Choose a detection method and its parameters in the Detect visual settings.
Shortcut Effect
KPI The step of the market equation shown on the chart. Each option shows its interest score.
Dimension A dimension to split the KPI by, or None for the total.
Segments The segments of that dimension to show. Appears once a dimension is chosen.
Denominator Show or hide the denominator of the step.
Main KPI Show or hide the total KPI, as a dashed line, when the equation has several steps.
Confidence interval Show or hide the confidence interval.

With several segments, right-click the chart, then Display, to Overlay them on one chart (default) or Split them into one chart each.

The Scores table lists every KPI, then its segments, sorted by interest score:

Column Content
KPI The step, and the dimension of the segment.
Segment The segment, and its absolute impact.
Outliers The number of outliers flagged.

Select a row to show that KPI or segment on the chart.

The comment describes the outliers in the visible part of the chart (KPI, segments, latest points), then lists the other areas of the Scores table, ranked by interest. See also Read the results.