I agree but they've also been positioning Looker/BigQuery as a more open data stack recently. Specifically trying to market Looker modeler (the semantic layer part of Looker) as its own stand alone tool

Yeah I've been following cube for a while now, lots of great new developments from their product, and love how they are breaking the metrics/semantic layer further out of the BI tools

I never understood this, it's literally just extra lines of SQL code... because *reasons*

There's so much value to add anywhere along the data chain.

Honestly, blending business context (analyst focus) with data modeling is an extremely impactful role.

Yes there's some shit AE models out there, but I've seen some terrible SQL written by data engineers too

I tried that for Google Analytics 4 issues and it didnt work 😀

2) Set up reverse ETL data feeds that power machine learning experiences on a website (fit predictor)

1) Designed and built an ecommerce data mart from scratch, integrating GA4 and GA UA data to provide continuous web analytics reporting before and after the Universal Analytics sunset

Integrate their product? Or buy the [future] competition?

how is your neck not absolutely wrecked?

Very often I will see heavy sampling.. like 2% of available data with GA4 explorations

We use the GA4 UI less and less.. and BigQuery / data mart is our "source of truth" for everything.

damn I've found my people

andddd that fixed our paid session count bug:

I fixed it with this little bit of code:

CASE
WHEN REGEXP_CONTAINS(page_location, r'gclid=.+') AND traffic_source_source = 'google' THEN 'cpc'
/* Known GA4 misattribution bug */
ELSE traffic_source_medium
END
AS traffic_source_medium

There were records where traffic source medium was "organic", even though that same hit contained a GCLID and the traffic_source name was 'google'

This should be impossible, as the GCLID would only populate on an ad (and therefore should be cpc/paid traffic source medium)