The Activation Event Taxonomy clinics made our warehouse less theatrical. We deleted two “success” events that fired on page load. Finance noticed the activation rate drop and, for once, believed the next chart.
Voices
Not a wall of five stars. A record of what the work actually did.
We ask alumni to name the module that mattered and the part that still nags. If you only want applause, this page will feel uneven on purpose.
★★★★☆
Qualitative Drop-Off Interviews was quieter than I expected — almost no slides. The consent script for Thailand research is what I still forward internally. Would have liked one more observed session in the room.
Ploy, Phuket. I took Signal Seat recordings while on parental leave. The drop-curve templates transferred; the live peer defence obviously did not. Treat recordings as a reference library, not a substitute for being argued with.
Experiment Design for Onboarding stopped us from shipping a colour test on a KYC step that was failing because of a vendor timeout. Mild reservation: the sample-size homework assumes more weekly traffic than a niche B2B console actually has. Kanya helped us resize, but the worksheet defaults still lean consumer.
Retention After First Value was the first time our CX and data people used the same week-three definition. I would not take it before Cohort Funnel Diagnostics; the overlap is real and the later course assumes you already retired vanity events.
Case note · consumer wallet, Bangkok
A wallet team arrived convinced that “tutorial completion” predicted funding. Cohort Funnel Diagnostics week three showed tutorial completers were simply people with fewer device constraints. After retiring that event and overlaying eight drop-off interviews, the largest leak was a hidden permission prompt on older Android builds common in their upcountry cohort.
They did not A/B the tutorial. They shipped a pre-prompt explainer and a fallback path. Day-7 funded accounts moved from a noisy 31% claim to a duller, more honest 37% among the Android slice they actually cared about. The board pack got shorter. That was the win.
Case note · B2B console, remote cohort
An industrial software team had a handsome onboarding checklist and a quiet graveyard of “invited admin” accounts. Interview overlay found that first value was not dashboard setup — it was exporting a single vendor list for a colleague who never logged in.
They redesigned first-hour success around that export and stopped counting checklist ticks. Activation looked worse for a month. Sales stopped promising a metric the product could not defend. We count that as completion, even though the chart was less flattering than the one they arrived with.