We have been building and running the AgileData Information Platform on Google Cloud for just under 8 years, a choice we have rarely regretted.
We have had a set core architectural principles from day one, a couple are:
we would use Google Cloud serverless services by default
we would reuse over build our own
Some of the Google Cloud services were a no brainer for us, Google BigQuery, Spanner, App Engine, Cloud Run etc.
But some of the core services we need to have were surprisingly missing 8 years ago from the Google Cloud environment, or were not truly serverless offerings, but more bolted on legacy products with a thin serverless wrapper.
For example the ability to search for data in BigQuery was missing for years, surprising given Google is a search company, but eventually it arrived.
A graph database was missing, we got told to use the Neo4j offering, the very definition of a serverless sheep in wolf clothing. And then we eventually got Spanner Graph and now BigQuery Graph.
For the Metrics Layer we got told to license the full fat Looker, again another legacy product dressed up as serverless.
For Search and Graph we could get away without using that type of capability, and just wait for it to arrive as a Google Cloud service.
But as you would expect a Metrics Layer is kinda important for a Data Platform and so we took a piecemeal approach and built little bits when we had to, but never built out the full capability.
We saw Malloy start in the Google ecosystem and thought great, a serverless Metrics Layer must be nigh, and then watched with dismay as its creator moved on somewhere else and Malloy left with him.
And so it was with great excitement we saw the emergence of BigQuery Graph and its Measures capability.
https://cloud.google.com/blog/products/data-analytics/bigquery-graphs-with-measures-for-trusted-agentic-workloads
https://cloud.google.com/blog/products/data-analytics/bigquery-graph-connecting-data-and-ai-at-scale
So as is the AgileData way we started a set of small McSpikeys (experiments) to see if we finally had something we could reuse as a fully functional Metrics Layer.
We had access to an early version and determined we could move our Measures there but not our Metrics (more on the definitions of these later), so we waited for the GA release.
While we sometimes get access to early release features as a Google Partner we are under strict NDA for those and so this series will only ever talk about features that are publicly available.
As part of the GA release we saw the ability to support Metrics, which was enough to start the work of migrating as much as we could to BigQuery Graph for our Metrics Layer capability.
And as part of that work I decided we might as well document our journey as we go.
As I have said many times before, writing helps me think, and there is a lot of complexity in the current cobbled together Metrics Layer we have built so far, so writing about the journey as we go on it actually helps me document and understand that complexity and hopefully gets us to more simplicity.
I will update this article with each new part of our journey so you can follow along from here.
Come follow our journey to Metrics Layer simplicity, you know you want to ;-)
The Journey
Part 1 - The Why (this article)

I've been watching the BigQuery metrics/measures capabilities intently and wondering.. what about the front-end visualizations of measures?
Yes, there is conversational analytics, but what about support or integration for legacy BI tools. For example, could we connect BigQuery graphs to Tableau and have the metrics live centrally in BigQuery?