A dimension is a characteristic that describes your data, such as the page someone visited, the city they were in, the device they used or the campaign that brought them. Dimensions are usually words or labels; the numbers you measure against them, such as sessions, enquiries or revenue, are called metrics.
How dimensions work
Every row in a report is defined by its dimensions. A table with “Session source / medium” as the dimension and “Sessions” and “Key events” as metrics gives one row per source, with the counts beside it. Add “Device category” as a second dimension and each source splits into desktop, mobile and tablet rows. Looker Studio colour-codes the difference: dimensions appear in green and metrics in blue.
GA4 collects many dimensions automatically, including page path, country, city, browser and the traffic source fields. Each has a scope that sets what it describes. Event-scoped dimensions describe a single action, such as the link that was clicked. Session-scoped ones describe a visit, such as the channel that started it. User-scoped ones describe a person, such as whether they hold a membership. Item-scoped ones describe a product within an ecommerce event.
When GA4 does not collect what you need, you can send it yourself as an event parameter or user property and register it as a custom dimension: the service an enquiry was about, for instance, or a customer’s membership tier. At the time of writing (October 2026), the limit on a standard GA4 property is 50 event-scoped and 25 user-scoped custom dimensions, and each one starts collecting only on the day you register it.
Why it matters
Dimensions are how a report answers the question “which”: which pages, which campaigns, which areas, which services. A London accountancy practice that wants to know whether tax return enquiries come from search or from referrals needs the enquiry type as a dimension. Without it, every form submission looks identical and the question cannot be answered.
Location dimensions need care in the UK. City and region are estimated from the visitor’s network connection, so they are approximate. A visitor in Croydon may simply appear as London, and people on mobile networks can show up in a city some distance from where they actually are. Use location dimensions to spot broad patterns, not to compare one borough with the next.
Common mistakes
- Combining dimensions and metrics of different scopes, such as a user property with an event count, and reading results that do not mean what they appear to.
- Registering values with thousands of unique entries, such as full URLs with tracking strings or order numbers, as dimensions. These high-cardinality dimensions push rows into an “(other)” bucket in standard reports.
- Expecting a new custom dimension to fill in past data, when it only records from the day it is created.
- Treating every (not set) row as a fault, when it often just means the dimension had no value for that row.
- Sending personal details, such as an email address, as a dimension value. That breaches Google’s terms and creates a UK GDPR problem.
How to act on it
List the questions your reports need to answer, then check that each one has a dimension behind it. If “which service was this enquiry about?” matters, make sure your form sends the service as an event parameter and register it as a custom dimension now, so the data starts building. Keep the values short and grouped, such as “bookkeeping” rather than a full page title.
When a report looks odd, check the scope of each dimension before doubting the data itself. Building reports around the right dimensions is part of my performance marketing work, so each channel’s cost per lead can be broken down by the things you actually make decisions about.