Along with the change from UA to GA4 comes an entirely new data model.  Dashboards will need to be updated, reports will have to be redrawn, websites and apps will have to setup tags for GA4. In short, everything is changing.

Below is a summary of reporting changes relevant to Analytics Canvas customers.  For an in-depth summary, read out post on the 4 ways to export data from GA4.

Google now offers to export your GA4 data to BigQuery for both Standard and 360 accounts, and we strongly encourage all GA4 property owners to enable to the BigQuery export for your GA4 properties.  

The data retention period for GA4 is at most 14 months. The only way to keep data longer than 14 months is to either store it yourself or enable the BigQuery export. 

The BigQuery export is not subject to the data limits and quota imposed by the Google Analytics Data API, however it is limited to 1M events per day unless you have a 360 account.  You can use the streaming export to get beyond 1M events in a free account, but now you're off the rails and not benefitting from any data reprocessing. 

Cost seems to be the biggest concern when it really shouldn't be.

The costs for storage are very low - from pennies per month for small sites, to dollars per month for most mid-sized sites. Incredibly large sites may pay a few hundred per month.

Processing costs will vary, but generally are very low.  With tools like Canvas you can set budgets and control your cost easily.  

If your site has any kind of meaningful volume or lots of events firing, please avoid the costly mistake of connecting your BI tools directly to the BQ Export. 

With your data in BigQuery, it can be imported into a Canvas workflow using the BigQuery connector in Canvas Desktop or Online, and you will be able to make queries that are just not possible with the API due to limits and quotas. 

Those with high data volumes will be impacted by the Analytics Property Quotas.  They are applied to the entire property both hourly and daily and everyone's usage counts towards it.  

There are also limits on features, such as Event Parameters and Custom Dimensions, based on Standard and 360 service levels. 

Requests for large numbers of rows, large numbers of columns, complex filters, or long date ranges will consume and potentially exhaust the per-hour and per-day quotas.

Analytics Canvas displays and reports the quota available and used by each of your queries so that you can see how your quota is being consumed by Canvas.

Canvas minimizes the amount of quota required by using a variety of techniques, including incrementally loading existing queries. Canvas also detects the more expensive queries and attempts to run the query in smaller batches or over longer periods to ensure a complete load. As a result, Canvas can often return a complete dataset for queries that fail in other applications. 

Collection Limits: 

There are a number of collection limits that may impact your reporting.  Unfortunately if you surpass a collection limit, the data will not be collected and therefore cannot be reported.

To manage within the 1M event export limit to BigQuery for Standard accounts, Google suggests excluding data streams and events during the linking process using filters. Of course you can always upgrade to Analytics360 for GA4, which does not have any limits on data collection.

There is a form of sampling in the new GA4 web interface in non-standard reports and Explorations when the query is complex or or exceeds 10M events for Standard accounts and 1B events for 360 accounts.

While there is no sampling in the GA4 API (horray!) if your query is too expensive, GA4 will just return NO data instead of sampled data (boo!).  

Similar to avoiding sampling with our UA connector, we have developed algorithms that detect these ‘expensive’ queries and respond to the API’s error messaging. Where possible, Canvas then develops a query plan and executes it in batches that are acceptable to the Data API, returning a complete data set to the user.

This means that we can often return to you the full data set in cases that would fail through the web UI or other API connectors.

High Cardinality Sampling - aka Report Query Limiting

Those who encountered "Report Query Limiting" in UA, where high cardinality columns resulted in a row of values called "(other)" will be familiar with this concept.  

Whereas in UA, standard accounts had a limit of 50,000 unique values in a column, in GA4 this number is further reduced to 20,000 unique values per day. However, this is not as straight forward as it sounds.

That 20K number is not based on your individual query, but rather the backend table that is housing the underlying data. For this reason you may see "(other)" even in cases where the cardinality is far less than 20,000.