Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Glossary term

Cardinality

The number of distinct values a dimension or field holds; high-cardinality dimensions, such as page paths with IDs, can exceed report row limits.
AnalyticsconceptUpdated

Definition

Cardinality is the number of distinct values in a dimension or field. Low-cardinality dimensions, such as country, device category or default channel group, have a short, stable list of values. High-cardinality dimensions, such as full page URLs, search terms, timestamps or customer and session identifiers, have values that are mostly unique.

High cardinality is not a fault in itself. A search-terms report is meant to be long. It becomes a problem when a tool's limits on rows are reached, at which point the tool must decide what to show and what to group, and analysis built on the full list quietly stops being possible.

In practice: which dimensions run high and what that does to reports

The guide to GA4 reports and dimensions explains the mechanism: each dimension has a documented daily cardinality limit, and once a day's distinct values exceed it the remainder is aggregated into a single "(other)" row that then appears in every report using that dimension. The usual causes recur across the site's GA4 guides:

  • Page paths carrying identifiers, such as session, order or product IDs in the query string.
  • Search terms, which the site search analysis tutorial normalises and analyses in explorations and the export for exactly this reason.
  • Custom dimensions holding unique values: timestamps, full URLs or free text, which the guide to GA4 events and parameters warns against registering.
  • Item lists named from page addresses or free text, which the merchandising analysis tutorial lists among its common mistakes.

The remedy runs in three directions. Reduce cardinality at the source: send a category, template or stable identifier instead of the raw value, and normalise free text before it is sent. Move the question to an exploration when it needs the long tail. And query the BigQuery export, which receives every event without cardinality limits, when exact detail is the point.

Cardinality, thresholds and sampling compared

Three mechanisms hide or alter data in GA4, and they are easy to mistake for one another. The common GA4 mistakes guide separates them by their signature.

MechanismWhat you seeCause
Cardinality limitA large "(other)" rowToo many distinct values in a dimension in a day
Data thresholdsMissing rows and a warning iconSmall rows withheld under a reporting identity that includes signals or modelled data
SamplingAn icon in an explorationThe query covers more events than the exploration's quota

"(not set)" is a fourth label that looks similar and is unrelated: it means the dimension had no value for those events at all.

Illustrative example — not a real result

An auction house stops logging every lot number as a dimension

Gavelbrook Auction House runs online sales of furniture, prints and silverware with about 3,000 new lots a week. To see which lots drew interest, it registered the lot number as an event-scoped custom dimension. Within a month the lot report was mostly "(other)": on busy sale days, 82% of lot views sat in that single row. The analyst replaced the dimension in reports with lot category, which has 24 values, and kept the lot number as an event parameter for the BigQuery export, where a weekly query lists the most-viewed individual lots. The category report has shown no "(other)" row since, and the specialists get their lot-level list from the export.

A Custom Dimension is where cardinality is most often created, because any registered parameter becomes a reporting dimension with whatever values it carries. An Event Parameter holds the value; choosing what goes into it is the design decision that sets cardinality. Data Sampling is a different way reports lose precision, by calculating from a subset of events rather than grouping rows. GA4 is where most marketers meet the "(other)" row, and its export is the usual way around it.

Common confusions

  • "(other)" is a traffic source or a page. It is a bucket of many values that did not fit, not one value.
  • The (other) row is a tracking fault. Tracking worked; the dimension holds more distinct values than the report can show.
  • "(other)" and "(not set)" mean the same thing. One is too many values; the other is no value.
  • A longer date range fixes it. The limit applies per day, so widening the range does not bring the tail back.
  • High cardinality should be avoided everywhere. Search terms and URLs are legitimately high-cardinality; the aim is to analyse them where limits do not apply, not to stop collecting them.

More terms to look up

Where this term is used

Sources

  1. [GA4] About events — Google Analytics Help

Related