GA4 and Databricks
If your business runs on Databricks, your GA4 data belongs there too. Marc Alexander brings every GA4 event into Delta tables, models it into sessions and conversions, and joins it to the rest of your data.
What you get.
- GA4's BigQuery export switched on and checked, in an EU or UK region
- Each day's events copied into Delta tables in Databricks, with the full nested structure
- A GA4 model: clean event, session, user and conversion tables built from the raw export
- Access through Unity Catalog, and costs kept down on both Google Cloud and Databricks
- Multi-touch attribution, with the models GA4 no longer offers
- Dashboards in Databricks, Power BI or Data Studio reading from the model
- Documentation, query templates and training on the GA4 data structure
As lead analytics consultant at Halfords, Marc built a multi-touch attribution model on Databricks: six rule-based models across 221 columns, refreshed weekly behind twelve quality checks and feeding Power BI.
Packages.
GA4 to Databricks connection
£500 to £1,200
The GA4 export landed in Databricks as Delta tables, refreshed daily, checked against GA4 and documented.
Databricks implementation
£1,500 to £3,500
Connection, a GA4 model of session, user and conversion tables, query templates, a dashboard and training.
Advanced Databricks
£3,500 to £8,000
GA4 joined with CRM, ad spend and orders, multi-touch attribution, and training.
Google and Databricks bill their usage to you separately. If BigQuery is your warehouse, see GA4 to BigQuery.
How it works.
Scope
Marc reviews your GA4 setup, your Databricks workspace and the questions you want answered, then confirms the route, package and expected running cost.
Export and land
He switches on GA4's BigQuery export, then sets up the daily copy into Delta tables and checks the numbers against GA4.
Model
He builds the session, user and conversion tables, and any joins to CRM, orders or ad spend in the package.
Train and hand over
Your team gets the documentation, query templates and a training session. Ongoing support is optional.
What do people ask about GA4 and Databricks?
Can GA4 export straight to Databricks?
No. GA4 only exports raw event data to BigQuery, so the export is switched on there first. From BigQuery, the data reaches Databricks by a daily copy into Delta tables, by querying BigQuery from Databricks, or through a pipeline tool. Marc recommends the route that suits your platform and budget.
Why not just query BigQuery from Databricks?
You can, with Databricks' Lakehouse Federation, and it's a quick way to look at the data. But GA4's event parameters are nested, and federation brings nested columns across as text. For reporting that's used every day, a copy in Delta tables keeps the full structure, runs faster and doesn't add BigQuery charges to every query.
Can you build multi-touch attribution in Databricks?
Yes. GA4 removed its first click, linear, time decay and position-based models in November 2023, leaving data-driven and last click. With every event in Databricks, rule-based models can be built and compared side by side, and joined to orders and ad spend. Marc built six rule-based models for Halfords.
What does it cost to run?
You pay Google for the BigQuery export and any queries that copy the data out, and Databricks for storage and compute. For most GA4 properties both are modest. Marc gives you an estimate for your data volumes before any work starts.
Does our team need to know SQL?
Not for everyday use: dashboards sit on top of the modelled tables. Anyone who wants to write their own queries gets the query templates and training, and the GA4 SQL in Databricks guide shows how to write common reports.
We already use BigQuery. Is this for us?
Only if your company's data lives in Databricks. If BigQuery is your warehouse, GA4 to BigQuery is the simpler choice. If the rest of your data, such as orders, CRM and finance, is in Databricks, bringing GA4 there means one place to join it all.
What about GDPR?
The GA4 export stays in an EU or UK region, and the Databricks workspace and storage can be in the same region. Access is controlled through Unity Catalog, with a retention policy if you need one. The pipeline and SQL are documented, so your own team can run them.
More detail: the routes from GA4 to Databricks
- A daily copy into Delta tables. BigQuery exports each day's table to Cloud Storage as Parquet, and Databricks loads it. The full structure survives, and the export itself isn't billed by Google, only the query that reads the data.
- Querying BigQuery from Databricks. Lakehouse Federation reads the export in place, with no copy to maintain. Good for exploring; nested columns arrive as text, and each query runs against BigQuery.
- A pipeline tool. Fivetran's GA4 Export connector reads the export and flattens event parameters into columns. Less to build, with a monthly fee.
The SQL and the pipeline are documented and version controlled, so anyone with SQL can maintain them. Ongoing work is available on a monthly retainer, from £500 a month.
Unsure your tracking is worth exporting? A GA4 mini audit checks it first.
Last updated
Free tool
What does your site fire before consent?
Paste a page address. The tracking checker shows what a first-time visitor triggers before agreeing to anything, ranked by how much it matters.
Related work.
Retail · Lead consultant
Halfords
Halfords' tracking had no single owner. Six months later, every tag was understood, the double counting explained, and server-side tagging ready for cutover.
Read the case studyAsk about GA4 and Databricks.
Tell Marc where your data lives today and what you want to answer that GA4 can't. He'll reply within one working day with the right route, the package and an estimate of the running cost.
GA4 and Databricks from £500.
Prefer to talk first? Choose "20-minute intro call" in the form.
Or email [email protected].