Serverless app development, Analytics, click and hiking trails
California
Joined April 2008
- Tweets587
- Following523
- Followers1.3K
- Likes554
I like using Claude Code for many tasks, but I feel Gemini has been giving me better results for SQL queries, particularly for BigQuery. Maybe Gemini has better knowledge there, which would not be surprising.
Just released a new community visualization that brings interactive radar charts to Looker Studio.Built with @observablehq/plot #LookerStudio
Check out the GitHub repo with full implementation details, sample data, and deployment instructions: github.com/mwhitaker/looker-…
Live demo: lookerstudio.google.com/repo…
No pagination in #GA4 explorations? Explorations are the only way to do proper segmentation, but you won't be able see more than 500 rows
Even if no data will be processed when GA UA shuts down, will this still be the case:
"If you have an existing configuration in the Google tag for your Universal Analytics property, it will be applied to your Google Analytics 4 property as well."
support.google.com/analytics…
One of the key ideas is analytics is "segment or die" and that data in aggregate is not good enough.
Boy don't they keep segmentation hidden in GA4. Can't apply segments to main reports and missing in API too. I guess everyone using BigQuery instead?
New Post. Cloud Run Jobs gets the job done for dbt and BigQuery.
michaelwhitaker.com/posts/20…
Quite impressed with my kid’s programming class. First order of business is to set up a Google Cloud project and create a http cloud function. There must be a gazillion “hello world” end points out there!
Very satisfying to be able to reproduce a report by the US census bureau from raw data. @observablehq is an amazing tool. observablehq.com/@monitus/us…
Trying out Vertex AI. Watch out for cost. Trained a model and it seems to run the full node-hours you specify. I assumed it would run a fraction of that since I used a small toy dataset. Ain't cheap if you just want to muck around and get into machine learning.
The simple pleasures in life. First pull request in my dbt GitHub Action 🥳.
I'll brush over the fact that it was due to an error in the code. github.com/mwhitaker/dbt-act…
Looking forward to the dbt conference #dbtcoalesce over the next few days. Love using dbt in my analytics workflows.
Who's ready for #dbtcoalesce? There are 4 sessions to meet the @fishtowndata team:
👉🏾 dbt 101 (EU & US-friendly)
👉🏾 Ask the experts: Customer product journey
👉🏾 Ask the experts: Version controlling your metrics
👉🏾 dbt 101 (US & AU-friendly)
Agenda here: hubs.li/H0C6sLk0
Workflows in GCP looks interesting, but I would have expected to be able to use the gcloud cli directly in a step rather than having to call low-level APIs via http requests. cloud.google.com/workflows/d…
Made a pitch to speak at getdbt.com/coalesce/
Even if there's no spot available it's already motivated me to work on some new material.
Running dbt with Github Actions in your repo is the coolest thing yet. That way you can have your application and data analysis all in one spot. I have barely started scratching the surface with the workflows you can create. github.com/marketplace/actio…
I think I finally understand BI Engine a bit better. As long as the SQL query returns less than the current 1GB limit for Data Studio (and as long as that query doesn't use some advanced operators) it will get accelerated by BI Engine. 1/3
I assumed you had to designate a dataset or table for use with BI Engine, but that is not the case. You can have as many tables in BigQuery as you want. The only thing that matters is that the size returned by the query is less than 1GB. 2/3