Case study · Bangkok Q2 2026
24 students. 8 weeks. +1.2 CEFR speaking levels.
Our first full cohort ran February–April 2026 in partnership with a Bangkok coworking community. Here's what we measured, what surprised us, and what we'd do differently.
- 24
- Enrolled students
- +1.2
- Avg. CEFR speaking gain
- 87%
- 8-week retention
- 4.6★
- End-of-cohort NPS
* Placeholder metrics — to be updated with verified results from the live cohort. All testimonials published with explicit written consent.
The context
The Q2 cohort was recruited through word-of-mouth in two Bangkok coworking spaces. All students were Thai gen Z (18–26), already working in tech or design, and rated their spoken English as "intermediate" in a pre-survey. None had used AI-first learning tools before.
| Metric | Baseline (week 0) | Exit (week 8) |
|---|---|---|
| Avg. speaking CEFR | B1.1 | B2.1 |
| Weekly active days | 2.1 | 4.8 |
| Coach Niramai sessions / week | — | 6.3 |
| Grammar error rate (IELTS proxy) | 34% | 18% |
Early voices
"Niramai never made me feel embarrassed for making mistakes. After three weeks I started speaking in meetings without rehearsing first."
P., 23 — UX designer, Bangkok
Published with explicit written consent. Name anonymised on request.
"The tonal feedback was what I needed. My previous teacher told me what to say; Niramai showed me why it sounded off — and replayed it."
T., 21 — junior developer, Chiang Mai (remote)
Published with explicit written consent.
"I went from avoiding English Zoom calls to leading one in week six. I didn't expect that so fast."
K., 25 — product manager, Bangkok
Published with explicit written consent. Role details generalised.
What we'd do differently
Honest retrospective — the things that didn't go as planned.
- Week 3 drop-off correlated with exam season — we'd time-gate the hardest mechanics.
- Grammar-only learners stalled; adding the speaking module in week 1 (not week 3) would have helped.
- Async voice notes are loved but mobile push wasn't live yet — daily nudges would lift the habit loop.
- Peer leaderboards worked better when cohort sizes matched. Mixed-level groups need bracket seeding.
Next intake opens Q3 2026 · Limited to 30 seats