“Signal Lab forced us to throw out a checkout flag that looked green only because the cache still served the old screen.”
Nattapong K. · growth lead, Bangkok fintech
Read the experiment, not the dashboard glow.
Cache Nodecore is a small Bangkok studio for app analytics. We sit with product, data, and growth leads who already run flags and A/B tests — and still argue about what the numbers meant after a release.
11
product organisations currently on the ledger in Thailand
8 weeks
standard studio cycle for Signal Lab
63
working sessions delivered across 2025
2 days
typical reply window from the Pranakhon desk
Programs
Courses built around real mobile traffic, not toy funnels.
Each program names a specific failure mode we keep seeing: late events, sticky caches, holdouts that were never truly held out.
Flagship · 8 weeks
Signal Lab
The studio residency for teams who already ship experiments and need a shared grammar for variants, guardrails, and release windows.
Intensive · 4 weeks
Cohort Grammar
A short desk on exposure logging, user-level assignment, and why “new users this week” is rarely a cohort.
From the room
What teams say after the first honest readout.
The Cohort Grammar desk was slower than I wanted. We spent two sessions just mapping exposure. That irritation was useful — our North Star had been counting people who never saw the variant.
Mira · product manager, Chiang Mai
★★★★☆
Clear on holdouts. Less magical on iOS ATT delays than I hoped, but they said so up front.
Client in grocery delivery · platform review
How we teach
A ledger, not a motivational funnel.
We do not promise a universal experimentation stack. We teach you to interrogate the one you already have: assignment logs, metric windows, cache TTLs, and the quiet ways a mobile client lies.
- Primary lensAssignment before metric
- Studio habitWrite the limitation first
- Classroom artefactOne shared readout sheet
- AftercareOffice hours in Bangkok time
Pranakhon desk
Taught from New Road, written for Thai release calendars.
Most of our cohorts sit inside organisations that ship on Thai public-holiday rhythms, mixed iOS/Android stores, and analytics vendors that were configured three PMs ago. The studio is small on purpose. We would rather keep eleven teams honest than scale a catalogue.
Journal
Recent notes from the ledger.
Method
Holdout groups without the leftover noise
Why a 5% holdout still contaminates if your cache key ignores assignment.
Release
When a variant is actually ready
A checklist we use before anyone is allowed to say “we can ship this.”
Thailand
Experiment cadence for Bangkok product teams
Store review lags, Songkran freezes, and other calendar facts dashboards ignore.
Desk hours
If your last experiment still has three interpretations, write to us.
No automated funnel. A coordinator in Pranakhon reads the note and suggests a program, a single working session, or a polite decline if we are the wrong room.