Names are a control surface
If anyone on the squad can invent user_email_raw, you do not have a taxonomy problem. You have an access problem dressed as analytics.
Cloudcacheline · Rayong
We teach analytics teams to treat schemas, consent flags, and retention clocks as owned work — not a backlog item for later. App analytics data governance is the craft; the warehouse is just where the mistakes accumulate.
18teaching cohorts since 2020
11Thai product organisations we have sat with
7 wksflagship stewardship path
6 hrslive desk review each cohort week
Flagship
A seven-week studio path for people who already ship events and are tired of unnamed properties, silent PII, and “temporary” tables that never expire. You leave with a contract template, an owner map, and a change-request ritual your warehouse team will actually honour.
Informational tuition: ฿48,500 per seat · teams of five from ฿186,000. No checkout on this site.
How the desk works
Most analytics training starts with charts. We start with who is allowed to invent a property name, and what happens when that name is wrong six months later.
01
You bring a real stream — even a messy one. We list events, owners, and downstream copies before we argue about taxonomy philosophy.
02
Each week a pair presents a one-page data contract. The room edits it out loud. Vanity metrics do not survive that exercise.
03
Retention clocks, consent flags, and access notes are written so a legal reviewer in Thailand can follow them without a decoder ring.
From recent seats
The owner-map from week two replaced our Notion graveyard. We still argue about naming, but at least we know who is allowed to rename
checkout_started.
I came for PDPA retention windows and got more value from the consent-flag lab. The warehouse lineage module felt rushed for our stack, so we booked a Resident Desk day afterwards.
★★★★☆ · Verified cohort seat · August 2025 — “Useful, occasionally slow. The change-request ritual stuck; the sampling lecture did not.”
What we insist on
If anyone on the squad can invent user_email_raw, you do not have a taxonomy problem. You have an access problem dressed as analytics.
A flag that dies at the first ETL step is decoration. We teach how to carry it into warehouse copies without inventing a second truth.
PDPA-aware windows only matter if someone can show when a table is due to die — including the “debug” extract on a laptop.
A living owner map, reviewed monthly, beats a beautiful Confluence tree that nobody updates after the reorg.
Journal
Product teams still encode phone fragments and ticket IDs in event titles. Here is how we catch that in week one.
A 90-day policy is fiction if the BI extract has no clock. A Rayong checklist for analytics stewards.
Why we retired the “single source of truth” page and started assigning humans to event families.
Visit or write
Tell us which stream is misbehaving. We reply within two working days and will not invent a checkout flow.