Understanding how the user metric is measured in Google Analytics (GA) is a more complex process than its straightforward definition. In layman’s terms, a user, is a person who has visited a website within a reported time frame. Where this understanding gets more complex is when reporting criteria is layered into that count in GA and forces the metric to be processed in the form of a pre-aggregated table or recalculated based on the dimensions being used. It is not as simple as a static counting metric like sessions, which is why so many people get confused as to why their reports can change drastically when comparing reports over different periods of time and with multiple dimensions.
Similar to the layman’s definition of a user, Google defines the user metric as the quantity of users who have viewed or interacted with your website or application. Unlike its metric relative, sessions, which is incrementally counted, users are deduplicated and either processed by table reference or recalculated depending on the type of report that is pulled. These two distinct measurement processes determine the results that are displayed and are described in detail below.
Referenced Data (Time Series)
If you are only using a time series dimension in your report like day, week, month or year, then the user metric is calculated by referencing a pre-aggregated sessions table. What this means is that if a user had at least one session within a the time series in the report, then the user will be counted. As you probably thought after reading that last sentence, this process is not too complex and takes less processing time to generate a report as it is references stored data rather than calculating based on conditions created by adding non-time based dimensions to the report.
Recalculated Reporting (Dimension Based)
Whenever you are creating a report that does not strictly use time series dimensions, then you will see varying user counts depending on how the metric is bucketed within the dimensions used in the report. It is important to note that in certain circumstances when an acquisition based dimension is used in a report that there will be overlap in the user count. This is due to the fact that a user can come through multiple acquisition channels within a chosen time period, which is why they will be counted more than once for individual channels.
Why Doesn’t the Total Count Match the Row Total?
As you read about the different user measurement you may have noticed one commonality between the screenshots, which was an identical total user count of 577. This is not by accident, as similar to how users are deduplicated for the dimensions being used, they are also deduplicated as a whole, which is represented in the total user count at the top of the reporting. This is a representation of users as a whole within the selected date range and not of users within the date range that fit the dimension criteria. The resulting effect that is created by this distinction is that the row totals will never add up to the total user count. As I mentioned in the Recalculated Reporting section, there are cases where users will fit the criteria of multiple buckets, which will inflate overall user numbers by including the same user in multiple rows, but will only be counted once in the total user count.
As you can see from this in-depth look at the user metric that it is actually a more complicated metric to analysis than one would have first thought. The process differs depending on the time and the dimensions being used, which has the potential to cause a lot of confusion if not fully understood. Remember this the next time you create a custom report and understand why counts will differ and what that means about the users that are acquired.