How much will the GA4 BigQuery Export cost?

There are two components of BigQuery cost:

1. Storage - the total storage used by your tables about $20 per TB per month ($0.02/gb/mo of active storage, and $0.016/gb/mo for long term storage in a specific region. Multi-region pricing is lower). The first 10GB of Active and 10GB of Long Term storage is free.  


2. Query Processing - the bytes that are processed by the queries you run, billed at $5 per TB queried per month.  Bytes are calculated by looking at the data stored in the columns used for the date range queried.

Storage is generally a very small cost.  The main driver for storage is how many events you have per day and how big those events are. 

You can find the size of your events in the BigQuery Console.  Click on the events_ group of tables, then under Details > Storage Info, you'll see both the number of rows and total bytes for the table.

After 90 days without an update, the daily tables are moved to long-term storage, which costs less. 

We estimate that on the very high-end, a single event will be 1,500 bytes.  With 1,000,000 events per day, that would be just $6.06/month in storage costs in the US multi-region location. 

Using the BigQuery Storage Calculator, enter your estimated amount Active Storage for 90 days, and Long-Term storage for 275 days to get your estimated monthly cost after 1 year. 

Using the storage calculation above and our conservative estimates for total events and event size,  if you have a GA4 property that records 50,000 events and 36.37  megabytes per day, after 90 days it will store 3.2 gigabytes.  Over 275 days it will hold 9.76GB.  After 1 year, the GA4 BigQuery export cost for storage is $0 (zero) per month.   

The bottom line is, storage of the raw GA4 data does not cost much and for many sites will cost nothing at all. 

Query processing is $5 per TB per month (multi-location), where the bytes are the total bytes stored in all the columns required within the query. If the table is partitioned or sharded, only the bytes in the partition or shards are counted.

Query processing is likely to be the larger factor but still reasonable. Using the same example for the underlying data, imagine you run 1000 queries a month and that each query accesses about 10% of your year's data. 

That means in the above example that about 6.6 TB would be queried, and it would cost about $30 a month for those 1000 queries.  

In other words, 1,000,000 events per day x 1500 bytes per event x 365 days of data x 10% of data used in each query x 1,000 queries in a month = 54,750,000,000,000 bytes.  This is 49.8 TB queried in the month x $5/TB in processing charges = $249/mo in query processing costs.  

Those queries can be made from anywhere, including the BigQuery console or your BI tools. Each query counts for every user (and every agency, and every consultant) who is hitting the raw tables.

Overall the cost is reasonable, however what happens when you have a dashboard connected directly to the raw tables with multiple charts and tables that is being accessed by multiple users, some of whom are interacting with the dashboard?  Well, potentially, each user is causing a full refresh and each interaction is causing even more queries.  This is where your costs will really start to add up!!  

There is a lot of uncertainty and fear around runaway BigQuery costs, and certainly when someone gets a huge unexpected bill, the story often spreads.  The most common reason for an overly large bill is an issue with repeated, high frequency queries involving all the history over the raw events tables.  Here are some other reasons for high BigQuery costs: 

• automation that is set to run too often
• a badly written script 
• a reporting tool that generates a lot of queries from very large tables for each person who views the report (ahem, Looker Studio)
• inexperienced users writing bad SQL 
• development activity that is querying the entire dataset instead of a small sample
• production reporting that is hitting the raw event tables 

BigQuery does have budget control features, which can limit any issues. Setting these up properly will ensure you avoid any large and unexpected costs.  

The number one way to minimize BigQuery costs is to make smaller summary tables and update them incrementally. The image below illustrates a typical example of querying the raw event tables vs an incrementally loaded summary table. 

Analytics Canvas gives you an easy to use, visual query builder for querying against the raw event tables.  It then lands those queries into summary tables, incrementally loads them, and makes them available to other users within your subscription. 

If you exclusively use Analytics Canvas for working with GA4 data in BigQuery, the your query processing costs are $0/mo.  This is because the query processing costs are directed to your Analytics Canvas subscription, where the costs are controlled and capped by your budget.

With all of the query optimization techniques native to Analytics Canvas, the majority of customers need only 1x the Cloud Quota plan at $99/mo.