Question: Dealing with multiple Google Ad and Google Analytic properties

[jonathan-moulds-sb]

Hi,

Firstly thanks for the amazing resource, this really is a great accelerator.

I do have a question on approaches or best practices when consolidating multiple Google Ad and Google Analytics tenancies/properties into a single Marketing Data Store?

It looks like the template has been set up for Google Ads to handle multiple datasets but for google analytics Is it better to consolidate the "events_" table in each export into a single view or set up seperate dataform loads for each property? Could this be handled through these templates?

Any suggestions or guidance on approach would be greatly appreciated.

Thanks

[kingman]

Hi Jonathan,

For handling multiple GA4 exports corresponding to multiple properties we recommend to change the
GA4 export parameter to become a list of multiple dataset similar to the ads_export_data parameter. Then in dataform code make changes to the GA4 related pipelines to take property_id as a input so that it creates separate datasets for each property, something in the style: marketing_ga4_base_prod_xxxx and marketing_ga4_v1_prod_xxxx and for all the feature engineering and ML pipelines in MAJ the property parameter need to be added so that it sources the correct underlying dataset for it's process.
It's not recommended to consolidate event data for multiple properties into single table or dataset, as the data should be used separately for each properties in all the MAJ pipelines.

[jonathan-moulds-sb]

Hi @kingman, thanks for the response and your comment on splitting the properties makes sense, though obviously that is a different pattern to what happens when there are multiple ad accounts. My question is what should happen at the feature store and then pipeline phase? Would we create seperate feeds for each GA4 account?

[kingman]

Hi @jonathan-moulds-sb, at feature store and pipeline there need to be an additional parameter that specifies current GA4 account in the pipeline run, that dictates the input dataset and output dataset of the pipeline jobs, this it makes sure that separated datasets are created for each of the properties, for instance feature_store_xxx, purchase_propensity_xxx, activation_xxx etc.