“The leakage audit is the module I quote in stand-up. I still think calibration week is one evening too short if your plots misbehave.”
Pim, Bangkok
Flagship lab
Cohort Signal Lab
Six Bangkok evenings for people who already have an event stream and a nervous stakeholder. You will not leave with a deployed feature store. You will leave able to defend a score.
Modules
-
Identity before metrics
Anonymous, logged-in, and device grains. Logout, reinstall, and family-shared phones. You draw the ledger before you touch a KPI.
-
Event hygiene under time pressure
A compressed taxonomy pass: aliases, zombie events, and the minimum dictionary a hazard model can survive on.
-
Leakage audit
Forbidden joins, post-outcome flags, and campaign features that only exist for people who already converted. The week most people go quiet.
-
Session gaps as a leading signal
Hazard-style thinking without pretending you have a biostatistics department. Personal baselines versus calendar shocks.
-
Calibration and the operating note
Reliability curves, threshold choice, and the one-pager that tells CRM what the score is not allowed to do.
-
Public critique
Each seat presents. Faculty and peers mark the note. No slide theatre; the document is the artefact.
Learning outcomes
- Describe identity states on your app without collapsing them into “the user.”
- Find at least one leakage path in a familiar notebook.
- Propose a threshold with an owner, a review date, and a forbidden use.
- Write an operating note short enough for a PM to read on the BTS.
Informational fee
THB 24,900 for the six-week evening cohort, including recordings for enrolled students only. No payment is taken on this site. Employer invoices are available. See fees and the refund policy.
Instructor
Waranya Srisuk previously led growth analytics at a Bangkok super-app. She still reviews scoring notes for two regional products and refuses to teach paid-UA bidding, which is outside this lab on purpose.
FAQ
Do I need SQL?
Yes, enough to write joins and window functions. Students without SQL will need extra time on week two; the lab does not pause for a SQL bootcamp. Spreadsheet-only seats are a poor fit.
Can I bring production data?
Only a slice your employer has cleared, and only after week two. We start on a studio extract so critique can be public inside the room.
What will this lab not give me?
It does not include a production feature-store, a deployed pipeline, or media-buying tactics. You leave with notebooks and a documented scoring pattern. Teams without a warehouse must adapt the SQL labs themselves; we do not provision infrastructure.
How large is the room?
Sixteen seats. If we cannot give written comments, we will not sell a seventeenth chair.
Reviews from this lab
★★★★☆
Waranya’s comments on my operating note were blunt. I wanted a gentler edit; I needed the blunt one. Identity ledger from week one stayed.
Client in digital banking
Prefer the catalogue?
Shorter clinics live on the course desk. If you only need taxonomy or leakage, do not buy the flagship out of politeness.