Skip to content

Metrics and dimensions

An analysis uses two kinds of columns:

  • metrics, quantitative columns that can be summed (sessions, orders, revenue). They build the market equation;
  • dimensions, columns that describe the rows (date, country, channel, device, product). They define the comparison and explain the variation.

See Data requirements to shape your data.

To explain a step, Datama scores every dimension with an interest score: how well its segments explain the variation of the step. The most interesting dimension is shown first; you can switch to another one from the charts.

To force a dimension on one step, or keep a dimension out of it, use Focus dimension and Excluded dimensions in the step settings.

The Dimension hierarchy panel organises your dimensions as an indented list that drives the drill-down:

  • level 0 dimensions are explored first;
  • inside a dimension, Datama then explores its children or its siblings, never its parents. For example, with Country under Zone, Datama explains a zone, then the countries of that zone.

To edit the hierarchy:

  • Move up and Move down change the order;
  • Indent makes a dimension the child of the one above, Outdent moves it back up a level;
  • Pin this dimension explores it first within its level, whatever its interest score. Pin one dimension per level.

Two options of Compare use the hierarchy:

Option Values Effect
Cascade mix effects along the hierarchy On or off (default) Measures the mix of each level against its parent segment, instead of the whole step, so the mix effects of a hierarchy add up: Zone mix + Country mix within the zone = Country mix. The performance is then read at the finest level. There is one chain per level 0 dimension, following the pinned child (or else the first one) at each level. Off: each dimension is analysed on its own.
Cluster Other at each parent On (default) or off With the cascade on: rebuilds the Other segment inside each parent, so each level keeps its own segments. Off: builds it once for the whole comparison. This is faster on a deep tree (products, SKUs), but a level whose segments are all small then becomes a single Other.

In the web app, the hierarchy can be pinned to the story, together with its grouping settings.