The Event Taxonomy Studio did not make our board less hungry for a weekly number. What changed was that I could finally say which number was a screen view leftover from 2022. The verb catalogue from week two is still taped above the kettle in the Leeds office.
Voices
Not every chart improved. Several arguments got shorter.
These notes come from people who sat a Network Packethub programme. We asked them to name the module they remember, and we kept the milder complaints.
★★★★☆
Helena’s refusal to look at our production Mixpanel project was frustrating in week one and correct by week six. We wanted a rescue. We got a marked PDF. I would still like a later module on warehouse tests; it is not in the current syllabus.
Client in insurance. The Session Attribution Clinic explained why our deferred deep links were creating a second user on Android 14. We had blamed creative. The clinic was four weeks and that was enough; a fifth would have been padding.
I joined Signal Desk because I could not make Tuesday afternoons. The recordings of the properties seminar are dry, which I mean as a compliment. I paused them and marked our own sheet. No live critique, so I still do not know if I over-deleted.
★★★★★
Week seven memo. She crossed out my first paragraph and wrote “state the window.” I still hear that when I open a doc. Rated on the Event Taxonomy Studio page as well.
The identity graph week moved faster than our data team could absorb; we had to replay the seminar twice. The naming conventions stuck anyway. I would not send a junior analyst alone.
Case notes
Two longer accounts
A grocery app and forty events that did no work
A London grocery product had 128 events, 40 of which had not fired in ninety days. The team sat Cohort Atelier in late 2025. They brought a redacted Mixpanel export. In week five the room deleted auto-tracked screen names that had been powering a “browse funnel” nobody in operations recognised.
After the studio they published a 61-event catalogue with owners. Support tickets about the weekly dashboard did not vanish, but the product lead stopped defending tiles that were empty. They did not hire us afterwards; they sent a second PM to the next intensive. That is the outcome we prefer.
A fintech retention curve that was a tracking hole
A payments app in Manchester showed a D7 cliff that marketing had been trying to spend its way out of. Two analysts took the Retention Curves Intensive. The truncated window on iOS backgrounding accounted for a large part of the cliff; a smaller part was a genuine drop after a fee-disclosure screen.
The intensive did not “fix retention.” It split the story so the fee screen could be redesigned on its own merits. One of the analysts later wrote that they still disagree with our caution about auto-tracked screens. We left that sentence in because it is true.