Definition:
A dimension is a descriptive attribute used to classify, group, filter or break down data in an analytics tool. Each dimension contains possible values: for example, the City dimension may contain London or Manchester, while the Device category dimension may distinguish between mobile, desktop and tablet.
Dimensions add context to measurements, but they are not limited to describing visitors. They may also refer to sessions, events, content, products, campaigns or devices. Their meaning and compatibility depend on each platform’s data model.
Table of contents
Dimensions and metrics
A dimension answers questions such as which type, from where, with which device or on which page. A metric expresses a quantity, such as sessions, events, purchases or revenue.
The two components are commonly combined in reports. If the metric is the number of sessions and the dimension is Source/medium, the analysis can show how many sessions correspond to google/organic, direct/none or other values. An aggregate metric can also be meaningful by itself; the dimension adds the breakdown required to compare categories.
In digital analytics, dimensions may appear in rows or columns, but they are also used in filters, comparisons, segments and conditions. Their function therefore does not depend on a fixed position in a report.
Dimension types and examples
Names and scopes vary between tools, although common groups include:
- User or context: country, language or account type when this data is available and can be used lawfully.
- Acquisition and session: source, medium, campaign or channel group that provides context for how a visit or acquisition occurred.
- Event and content: event name, page path, screen title or product category associated with an interaction.
- Technology: device category, operating system or browser used.
The same name can have different meanings depending on its scope. For example, first-user source, session source and the source attributed to a key event do not necessarily describe the same point in time or produce the same breakdown.
How to use a dimension in analysis
Selecting a dimension requires connecting it to the question being investigated and to a compatible metric. A basic process can include:
- Define the question: establish whether the objective is to compare channels, content, devices, locations or another characteristic.
- Select a dimension and metric: determine which attribute provides the breakdown and which quantity will support the comparison.
- Set the scope: review the period, filters, data scope and attribution rules before comparing results.
- Validate collection: confirm that values arrive correctly, follow consistent naming and do not mix incompatible categories.
In Google Analytics, field availability and compatibility can be checked in its official catalogue of dimensions and metrics. Other platforms use different models, names and restrictions.
Predefined and custom dimensions
Predefined dimensions belong to the tool’s data model and are populated when the implementation provides the required information. Not all of them will be available in every report, property or situation.
A custom dimension adds a classification that the platform does not provide as standard. It can be used to analyse account type, form type or a product characteristic, for example. Its creation requires a defined purpose, name, scope and values, as well as configured and validated data collection.
Creating a custom field does not automatically turn any information into a useful dimension. It is important to check whether an equivalent predefined dimension already exists and whether the new classification addresses a stable analytical need.
Limitations when interpreting dimensions
A breakdown can be incomplete or misleading if its collection and processing conditions are ignored. Important limitations include:
- Scope and compatibility: not every dimension can be combined with every metric, and an incorrect combination may produce inconsistent results.
- Cardinality: a dimension with too many distinct values may group categories, increase noise or make a report harder to interpret.
- Missing or modelled data: consent, data thresholds, configuration and technical limitations may prevent all values from appearing.
- Quality and privacy: tagging errors, naming changes or improperly collected personal data reduce reliability and may breach applicable rules.
A dimension organises the context of data; it does not by itself explain why a result occurred. Interpretation needs to consider the field’s definition, scope, implementation quality and the metrics with which it is combined.