Event Taxonomy Studio
A four-evening rewrite of event names, properties, and owner rules so downstream segments stop drifting.
App analytics studio · Bangkok
Segments that behave, not labels that sit still.
Devpacketline trains product and analytics teams to build behavioral segmentation for apps: cohorts that move when the product moves, event names that still mean something after a rebrand, and critiques that catch the quiet leak before the weekly review.
We work from a small studio on Ramintra Road. Classes are live, small, and deliberately slower than a conference talk. You leave with a segment library you can defend in a roadmap meeting.
- 27
- Bangkok studio evenings held since 2023
- 14
- Partner apps in the living case library
- 8 wks
- Typical span from first event audit to a segment that ships
- 6
- Faculty with shipping work on SEA consumer apps
Flagship
Cohort Signal Lab
The lab is our longest program. You rebuild one production segment from raw events, then defend it against faculty who have watched those same charts fail in Thailand’s super-app and fintech stacks.
Modules cover recency windows, identity after ATT, and the difference between a persona slide and a cohort that actually returns. Pricing is listed as information only.
How the desk works
Critique before dashboards
- Name what you actually capture Most teams inherit event names from a sprint three years ago. We start by reading the payload, not the slide.
- Frame a cohort that can fail A useful segment can empty out. If it never shrinks, it is a vanity bucket, not a behavioral cut.
- Ship one activation, then stop The studio asks for a single product response per segment. More than that usually means the definition is still mushy.
Programs
Short studios alongside the lab
Retention Arc Mapping
Plot return curves by first-week behavior instead of calendar install date, then pick one leak to close.
Funnel Fracture Clinic
Find the step where a “healthy” funnel is actually two populations colliding. Bring your own screens.
Case vignette
Grocery delivery, Bangkok
A partner team treated “weekly active” as one segment. After the lab, they split basket builders from coupon hunters using first-session add-to-cart timing. Promo spend dropped on the second group without touching the first.
The work took six weeks, not because the SQL was hard, but because product, growth, and data had to agree on what “hunter” meant in the event stream.
“Cohort Signal Lab forced us to throw out a segment we had been reporting to leadership for a year. Painful, and correct.”
The taxonomy evenings were slower than I wanted. We still finished with event owners, which our last vendor workshop never produced.
Journal
Recent field notes
Why session counts still mislead product teams
A session is a container, not a behavior. Here is how Thai-market apps inflate both.
Building a segment that survives iOS ATT
Identity is thinner now. Behavioral cuts have to stand without a perfect user id.
Bangkok studio
Write before you enroll
Tell us which app surface you own and which chart currently embarrasses you. We reply within two working days with whether a lab seat or a shorter studio fits.