About
A small analytics school that still meets in a room.
Systemtracebase opened so mobile teams in Thailand would not have to import a Silicon Valley playbook that ignores App Store queues, Thai festival traffic, and the way family-shared devices break naive user IDs.
Mission
We exist to raise the floor on Mobile A/B Test Analytics: honest hypotheses, defensible samples, and readouts that a release manager can act on by Friday. We are not a consultancy that ships your variants for you, and we are not a tool vendor.
The studio sits in Bangkhunprom, Pranakhon. Cohorts are bilingual in practice — materials are English, hallway talk often slips into Thai — and remote seats join from across Southeast Asia.
Pedagogy
Every week has three beats: a short lecture, a desk critique, and a written memo. Lectures stay under forty minutes. Critique is the expensive part. You present a live or historical mobile experiment; two peers and one instructor annotate the randomisation unit, the primary metric, and the first guardrail that could veto a ship.
We refuse “inspiration decks.” If a slide cannot be used inside an actual Jira ticket or experiment brief, it does not enter the room. Homework is short on purpose so people with release duty can finish it.
Community
Alumni keep a reading circle that meets on the first Thursday evening of each month, in person when possible. There is no points system and no leaderboard. People bring a failed test as often as a win; that is the point of the circle.
Student journey
Most students arrive mid-career: a PM who inherited Firebase Remote Config, an analyst who has never touched Store listing experiments, or a founder who has been peeking daily. The first two weeks diagnose instrumentation. Weeks three to eight design and run a contained test (often on a staging cohort or a low-risk surface). Weeks nine to eleven write the readout and a “what we will not change” list.
If your app cannot run any experiment during the cohort — for legal, store, or traffic reasons — we pair you with a historical case from the archive. You still write the memo. You do not fake a live test.