Apinya Sutham
Studio director · Signal Lab lead
Former instrumentation lead at a Bangkok super-app. Apinya designed Signal Lab after watching three consecutive “onboarding wins” collapse once the CDN TTL was shortened.
Studio
Cache Nodecore started as a reading group for Bangkok product analysts who were tired of weekly “wins” that vanished after a store release. The name is literal: node-level telemetry and client caches kept rewriting the story after assignment.
Teach mobile product organisations in Thailand to treat experimentation as a record, not a performance. A good readout names what the test cannot claim. A poor readout dresses a cache artefact as a conversion lift.
We work with teams who already have flags, analytics vendors, and a backlog of disputed charts. We are not a replacement for those tools. We are the room where the tools are cross-examined.
Every session begins with an artefact from the team: an exposure log, a metric window, a screenshot of a “significant” dashboard. Fellows annotate in silence first, then argue in a fixed order — assignment, observation, limitation.
We do not grade enthusiasm. We grade whether the limitation was written before the headline. That habit travels back into stand-ups more reliably than a new vendor.
People
Studio director · Signal Lab lead
Former instrumentation lead at a Bangkok super-app. Apinya designed Signal Lab after watching three consecutive “onboarding wins” collapse once the CDN TTL was shortened.
Metrics tutor · Cohort Grammar
Daniel came from a payments analytics desk. He is the person who will stop a session if the denominator quietly changed mid-test. Fellows find this either clarifying or mildly infuriating.
A typical path starts with a written note to the desk, a short diagnostic call, then either Cohort Grammar or Signal Lab. After the cycle, alumni join a monthly ledger hour — not a networking mixer, a shared reading of one anonymised experiment from a Thai release calendar.
We keep the room small. Eleven organisations is already a full ledger. If we decline a cohort, it is usually because the team does not yet log exposure at user level, and we would be teaching theatre.