The Survival Curves lab was the first time I trusted a plot enough to show it in a board pack. Module nine still feels too opinionated on paywall delay — our CFO ignored it — but the Day-23 framing stuck.
Flagship programme
Cohort Signals: Predicting Mobile Churn Before It Happens
Twelve morning sessions for analysts who are done decorating dashboards. You will leave with a named risk window, a feature store you can defend, and a memo written for a sceptical PM.
Informational tuition: THB 46,000 per seat, invoiced after placement. Remote and Bangkok room seats share the same cohort. Recordings remain available for fourteen days after each session.
At a glance
- 12 live sessions, 2.5 hours each
- Scored feature-store notebook
- One office hour in the following month
- Archive access during class only
- English instruction, Thai examples welcome
Modules
- 01 · The vanity funnel autopsy Why install-to-D1 charts flatter onboarding and hide silent decay. You bring one public screenshot from your own product; we take it apart together.
- 02 · Event names that survive a rebrand Intent versus tap. Students rewrite a 200-event dictionary down to a set that a new hire can learn in a week.
- 03 · Cohort geometry Calendar time versus user time. Thai public holidays and payday cycles as confounders, not footnotes.
- 04 · Session depth as a leading signal Screen sequences, widget use, background refresh. The module where most people discover their “active user” definition is theatre.
- 05 · Feature store, first pass Columns you are allowed to keep. Leakage hunts. A scored review, not a participation trophy.
- 06 · Survival Curves lab Kaplan–Meier without mystique. Censoring explained with prepaid SIMs that vanish mid-week.
- 07 · Hazard a PM can argue with Cox in hallway sentences. If you cannot say it standing up, it does not ship.
- 08 · Store-review language (TH/EN) Sentiment as a weak but early cue. Bias, sarcasm, and why one-star spikes after an update are not always churn.
- 09 · Paywall and prompt timing A joint session with the shorter Paywall desk. You will likely recommend delay. Marketing will not clap.
- 10 · Push fatigue and quiet uninstalls Campaigns that steepen hazard. Students delete or pause a live journey as homework when they have permission.
- 11 · The memo One page to a fictional CPO. Faculty grade the claim, the window, and the limitation — especially the limitation.
- 12 · Defence You present to a hostile PM role. The archive is closed. Notes only.
Learning outcomes
- State a risk window in days, not in slogans.
- Defend a feature store against leakage and vanity columns.
- Read a survival plot aloud without borrowing a data scientist.
- Separate review sentiment from session-depth decay.
- Write a limitation that finance cannot later claim you hid.
What this will not do
Cohort Signals will not invent demand. If the product is unwanted, prediction only tells you how fast people leave. We say this in session one and again in the memo rubric. Students who hoped for a growth hack usually ask for a refund before module four; see the refund policy.
Instructor
Kamonchanok Prasert
Former analytics lead at a Bangkok super-app, now faculty director. She still writes SQL in class and still refuses to publish a “health score.” Guest sessions from Niran Vong cover gaming economies; Kamonchanok owns the flagship arc.
Informational pricing
THB 46,000 per seat. Team of three or more: 8% courtesy on the same invoice, still no online checkout. Living Model retainers are listed on pricing.
FAQ
Do I need Python?
No. Notebooks are provided in a locked classroom environment. You may export SQL. If you already live in Python, nobody will stop you, but the grade is the memo, not the gist.
Can I join from outside Thailand?
Yes. Remote seats join the Bangkok morning slot (09:30 ICT). We will not run a second evening cohort this year.
Is there a real limitation you should know before paying?
Yes. Our labelled archive is strongest on commerce, mobility, and mid-core games in SEA. If your product is a regulated health app with sparse events, the archive will not resemble your world. We can still teach the method; we cannot pretend the examples match. That mismatch has frustrated past students in insurtech, and we would rather you know now.
Are recordings a substitute for live sessions?
No. Miss more than two live sessions without notice and we withhold the completion note. Recordings are for review, not for silent attendance.
From recent seats
★★★★☆
“Feature store week was brutal and fair. I wanted more OEM quirk coverage. Would sit it again for the memo rubric alone.”