Cohort Signal Lab
Rewrite event names, choose pedagogy-shaped windows, and present drop-off between lesson n and n+1 without blaming the learner.
Open the syllabusDatafeaturebase EdTech App Analytics · Thailand
Maps, not dashboards, for the way people actually learn.
We sit with product teams until lesson sequences, first-session value, and assessment events stop contradicting each other. The desk is small on purpose: one live app per cohort, exported tables you already own, and language a curriculum lead can argue with.
Next Cohort Signal Lab desk: 8 September, remote with optional days on Poochaosamingprai Rd. See the programme calendar.
Open desks
Three programmes currently taking seats
Each course is built around a live learning product, not a toy dataset. Browse all five on the courses page; these are the ones teams ask for first.
Activation Pathways
Instrument first-session to first-value for apps that teach, not shop. Useful when onboarding still looks like a consumer checklist.
View on the listing
Retention Cartography
Draw the actual path through a curriculum, including the quiet weeks after a mock exam, and decide what is product versus calendar.
View on the listing
Why the desk sits here
Thai learning products taught us not to import ecommerce funnels.
Tuition cycles, exam calendars, and family devices change what “weekly active” even means. A language app used on the BTS after school is not a SaaS trial. We started Datafeaturebase after watching campus LMS rollouts report green charts while lecturers quietly printed worksheets.
EdTech App Analytics, as we practice it, is closer to cartography than to growth hacking. You name the terrain (lesson, attempt, hint, skip, revisit), you choose a window that matches how the subject is taught, and you refuse metrics that only look healthy because the denominator is wrong.
Sixty-one product teams have sat with us since the desk opened. Some are Bangkok studios; some are regional teams shipping into Thai stores. All of them arrived with a dashboard and left with a smaller, meaner set of questions.
From recent desks
What people say after the event names get cleaned
We had been counting “lesson complete” whenever the video hit 90%. After Cohort Signal Lab, complete means the in-app check was submitted. Week-2 return looked worse for a month. That was the honest number.
Nalinee K., product lead Language app, Bangkok
Activation Pathways did not make our first session prettier. It made the third screen quieter, which is what our tutors had been asking for.
Client in K–12 homework tools Chiang Mai
The lab slowed our sprint for two weeks while we renamed events. I still think that pause was overdue, though I would not run it in the same month as a store listing redesign.
Preecha Hat Yai
How the desk works
Three habits we will not drop for a prettier chart.
-
Pedagogy sets the window
A cohort is not “everyone who installed in March.” It is everyone who started Unit 2 in the same teaching week. If your subject has a mock exam, that week belongs on the map.
-
One live product, not a sandbox
Seats include review of a real event table. We do not invent a fictional academy. If your warehouse is a mess, that is the curriculum.
-
Language a curriculum lead can veto
Findings are written so an academic partner can disagree in plain English. If only the data team understands the slide, it is not finished.
Field notes
Recent writing from the desk
Retention signals that survive a Thai exam calendar
Why week-4 return collapses in February for some tutoring apps, and which events still tell you something true.
Read the note
When completion rates mislead product teams
A completion metric that rises while tutors complain is not a win. How we separate video end from evidence of learning.
Read the noteWrite to the desk
If your funnel and your teachers disagree, that is enough reason to talk.
Tell us which app, which age band, and which metric you no longer trust. We reply within two working days from Samut Prakan. Programmes are informational; there is no checkout on this site.