Studio
A small practice on the west bank of the Chao Phraya
Cloudpacketbase began in 2015 as a reading group for researchers who were tired of treating session duration as a proxy for care. It is now a teaching studio for Session Quality Analytics, still based in Bangphlat, still suspicious of single-number stories.
Mission
We exist so product organisations in Thailand and across Southeast Asia can measure whether a session was usable, complete, and respectful — without laundering those ideas into a marketing metric. Our work sits between research, observability, and product operations.
The studio is deliberately small. Faculty rotate between teaching and field work with partner squads. That keeps the curriculum attached to production systems rather than to slide libraries.
Pedagogy
Every Cloudpacketbase course begins with a dataset, not a lecture. Students write instrumentation notes before they write scores. They present findings to a sceptical reviewer — often a visiting engineer from a Bangkok payments or logistics team — and they are marked on whether their caveats are specific.
We favour short written artefacts: a one-page score definition, a sampling log, a limitation paragraph. If a student cannot explain a number in plain English, the number does not ship.
Thai is welcome in discussion; course materials and assessed work are in English so mixed teams can share them. Evening cohorts follow Bangkok office hours. Daytime Field Lab sessions use the Bangphlat rooms when a dataset cannot leave a partner’s laptop.
Community
Alumni stay in a quiet working group: monthly critique of anonymised traces, no vendor pitches. Membership is included for twelve months after a live cohort and can be extended. We do not run leaderboards. We do publish, with permission, short field notes in the Journal.
Student journey
Most people arrive after a painful quarter: a replay tool was purchased, a “quality index” was launched, and nobody trusts it. They typically start with Instrumented Session Quality, add Latency Budgets if platform work is the bottleneck, then return for Cohort Replay Analysis once sampling is the unsolved problem.
A smaller group comes from research or CX and needs language that engineering will accept. For them we emphasise event contracts, privacy constraints, and how to say “this clip is not representative” without losing the room.
If you are unsure where to enter, write to us. We reply with a recommended sequence, not a package upsell. Fees are listed on the pricing page and are informational only.