Thailand
Retention signals we keep seeing in Thai consumer apps
Identity Flowcore journal · notes from the SEA Retention Clinic
Global playbooks treat retention as a smooth decay curve. Thai consumer apps rarely cooperate. The calendar is lumpy, the second screen is often LINE, and payment identity is frequently PromptPay rather than a stored card. If your model was born on a US grocery tutorial, it will misread Bangkok.
PromptPay retries are not “intent”
A cluster of failed PromptPay attempts followed by success is a settlement story, not a loyalty story. Teams that feed retry counts into a churn classifier sometimes invent a segment of “highly engaged at-risk users” who were simply fighting a bank timeout. We separate payment_failed from session quality before any score is allowed to see it.
LINE is a parallel operating system
Hand-offs to LINE OA or LIFF views look like drop-off in the primary SDK. They are often continuation. If you cannot join those sessions — and many teams cannot, legally or technically — the honest move is to mark the gap as “unknown channel,” not as churn. Pretending you observed the full journey is how lifecycle over-messages people who already finished the task in chat.
Festivals break personal baselines
Songkran, New Year, and Chinese New Year create legitimate silence. A hazard model that ignores the national calendar will chase users who are at home with family and then learn the wrong lesson when they return. In clinic we keep a simple overlay: do not treat a gap as anomalous if it falls inside a published holiday window unless it exceeds that person’s own festival-year history.
What we do not claim
These are recurring patterns, not laws. A B2B productivity app in Asok will not look like a wallet in Isaan. The point of teaching from Pathum Wan is to keep the texture available, not to sell “Thailand insights” as a slide.
If this is the problem you actually have, the SEA Retention Clinic is the shorter room. The flagship lab covers it only as context.