What is “(other)” in GA4?

(Other) is seen in the GA4 user interface in the Reports area, and in responses from the Data API that are using the Reports module as their data source.

This is because GA4 has limits of how many rows can be stored in daily tables used by Reports and by the Data API.  These row limits reduce the cost of processing for GA4.  

The row limit for many reporting tables is 50,000 rows.  And as Google says in the documentation, “When data collected in a day on one of these tables exceeds the row limits, the excess rows are rolled up under (other).”(1)

While to our knowledge the exact process has not been documented publicly by Google, it is clear that unlike UA, where the top 50,000 unique values had exact counts, and the long tail values were included in (other), it appears that GA4 will return different values for even the Top n most common dimensional values.

It's possible to see "(Other)" for reports with low cardinality columns - in other words, columns where there are just a few unique values for the dimensions you have included.

Unfortunately, it's not what is rendered, it’s what is in the GA4 behind the scenes table that had to be accessed.  If your results require GA4 to include a table that has been limited to 50,000 rows, you will see "(Other)" in the report and the values will be inaccurate.(2)

The result is that Explorer and Reports will often not match if Reports is showing an "(Other)" category. 

In these cases Explorer can be taken as the “right” answer, provided the results are not sampled.  Sampling kicks in for standard accounts when there are 10 million events in the result, and kicks in for Analytics360 accounts when there are more than 15 Billion events. 

You might have noticed that when you switch to the Explorer, you don’t see “(Other)” anymore, but you do see that your top 10 pages are not in the same order as they are in Reports.  How can it be that the order of the top 10 pages is different between reports and Explorer, just one menu item apart in the GA4 user interface!?

So we just learned that the one place you can avoid the dreaded Other is by using the Explorer. This will work for manual analysis, and providing you are not looking at more than the sampling limit.

To reduce how often (other) occurs in reporting you can also reduce the cardinality of all your dimensions.  You can do this by not creating custom events that have lots of unique values, for example a unique userID or a time stamp.  Consider removing all the query parameters from your page path before you store it in GA4. Otherwise you will see (Other) constantly with anything around page tracking even with medium traffic sites.

These techniques will help, but for a high traffic site, or sites with many pages or large numbers of unique products, they will not eliminate “(Other)”.  What can be done in this case? 

The only way to never have “(Other)”, and to analyze large numbers of events (past the sampling limits of Explorer) is to use the BigQuery export.

The BigQuery export provides access to the raw event data.  It does not include Google Signals data, so for things like demographics and interests, you will need to use Explorer or the Data API.

It also does not include calculated metrics and dimensions.  To get landingPagePath you will need to write specific SQL to find the page related to the session_start event, for example.

When using the BigQuery export table, it’s important to consider data volume and frequency of queries, since BigQuery is billed based on Bytes Processed.  You want to avoid unnecessary queries or repeatedly querying large tables to make summaries that you could query just once and then reuse.

Analytics Canvas allows you to connect directly to your GA4 BigQuery export tables, and also makes it easy to optimize your BigQuery processing and lower your BigQuery costs.

Canvas gives analysts the ability to quickly make datasets based on the BigQuery export tables at the level of detail required. It then delivers it directly to cost-optimized BigQuery tables or to tools like Data Studio and Sheets.  

It does this without constantly re-processing the highly detailed GA4 event data, thereby saving on BigQuery processing costs.

In the example shown above, simply connecting Data studio directly to the BigQuery event tables means that every time a user changes a filter or date range, there will be a query for every chart or scorecard on the report.  This means many queries on the detailed events tables, which are the largest and most expensive to query. 

Using Analytics Canvas, once a day, after the BigQuery export table is updated, Canvas updates a smaller summary table, with only the data you need for Data Studio or your own data-mart.  This table might be hundreds of times smaller than the raw table, so you can reduce your BigQuery cost significantly.