Flagship
Holdout Design for Product Retention
Six modules on building holdouts that survive seasonality, Thai holiday calendars, and a sceptical VP. Informational fee: THB 18,400.
Bangkok studio · since the 2019 table
Measure return, not theatre.
Logicfeaturegrid teaches Retention Experiment Analytics as a craft: how to build holdouts, read survival curves without overclaiming, and write a memo a product lead will actually sign.
Open the flagship bench
Who sits at the table
Marketplace, wallet, and media teams who already have events — and still argue about Day 7.
We do not teach SQL from zero. We teach the experimental layer that sits on top of a working warehouse: cohort construction, holdout hygiene, and the politics of shipping a kill decision.
6 Bangkok teaching cohorts since we opened the Chakrphong room
38 alumni notes collected after the 2024–25 benches
4.2 internal teaching score from those 38 notes — not a marketplace rating
1 studio. No franchise, no recorded-only factory
Open benches
Courses currently on the grid
Instrumentation
Cohort Instrumentation Studio
Name the events that actually define return. Built for teams whose tracking plan still treats “open app” as a north star.
Lifecycle
Lifecycle Messaging Experiments
Push, in-app, and email tests that respect a holdout instead of drowning it. Includes a module on quiet hours in Thailand.
“The holdout module stopped our team from celebrating a Day-7 spike that vanished once Songkran traffic left the sample. I still wish the SQL appendix went deeper — we had to pair a warehouse engineer for week three.”
“Short, slightly dry, and the only course that made me rewrite our experiment brief template. Rated 4/5 on our internal wiki, mainly because the live table runs late.”
How the room works
A table, a warehouse extract, and a kill criterion before anyone opens a chart.
Each live seat arrives with a real question from their product. We do not invent toy apps. You leave with a signed decision rule: what would make you ship, wait, or stop.
Homework is reviewed within seventy-two hours by the same instructor who sat with you, not a rotating assistant pool. That is why the table stays at eleven.
Read how the studio teachesWhat you take home
Three artefacts, not a certificate wall.
A holdout map
Who is in, who is out, what leakage looks like on your surface, and how Thai public holidays distort a four-week window.
A survival reading
How to describe a curve without claiming causality you did not earn. Includes language we have seen survive legal and brand review.
A decision memo
One page: hypothesis, unit of randomisation, primary metric, guardrails, and the pre-registered stop rule. No slide theatre.
From the journal
Field notes, not recaps of other people’s blogs.
Survival curves without the theatre
Why Kaplan–Meier on product return is useful, and where it quietly lies when users churn by uninstall rather than silence.
Day-7 windows and Thai marketplaces
A week is not a natural unit when payday, tourism, and monsoon delivery delays all move the same metric.