What this is
Your Supabase project already knows everything about your product. Who signed up, what they touched, when they stopped. Getting that out of Postgres and into a shape somebody can answer questions with is a different job, and most of the writing about it is either a vendor diagram or a tutorial that stops at “and now you have data in your warehouse.”
So we are building ours in public, on camera, one layer a week. Supabase Postgres, Supabase Pipelines into BigQuery, models on top, and the numbers our own go-to-market people ask for on Mondays. You get the working parts and the parts where we got it wrong.
Eight episodes after this one. Episodes 1 and 2 are out; the rest land weekly.
Who it is for
You run a product on Supabase and someone has started asking for numbers you cannot produce. Or you are weighing a warehouse and want to see the constraints before you commit to one. Or you already have the pipeline and the numbers still feel untrustworthy.
If you already run a modelled warehouse with a tested collection layer, this will be too basic. Skip to episode 8, which is the one about what breaks.
What you will have at the end
A publication you wrote on purpose, replicating into BigQuery. Models that answer activation, account health and expansion. And a clear read on which migrations break the chain, plus where to catch them.
The shape of it
Where it really starts
The pull request that changes the schema
Outside every guarantee below. Nothing downstream is built to question it.
Layer 1
Supabase Postgres
The record of what happened
Layer 2
Supabase Pipelines
Moves it faithfully, no silent loss
Layer 3
BigQuery
Models it into answerable questions
Layer 4
Your GTM tools
Acts on the answer
Why we are doing this on camera
Two honest reasons. The first is that we needed the stack. We build Skene, which reviews pull requests for tracking that got renamed, dropped, or never written in the first place, and we could not answer basic questions about our own product because our own collection layer had gaps. Building it in public was cheaper than building it twice.
The second is that this stack is new enough that the write-ups do not exist yet. Pipelines is in private alpha. The BigQuery destination has real constraints (primary keys, replica identity, no type changes) that you find out about by hitting them. If watching us hit them saves you an afternoon, the series has done its job.
