Voices
We asked alumni to mention a module by name and to keep at least one complaint. Several did without being asked.
The property dictionary from week two of Event Taxonomy Atelier is now in our repo, next to the schema tests. QA fails a build if plan_tier arrives as free text. I would still like a faster way to onboard contractors; the governance ritual assumes people will read the log, and ours do not always.
Helena V. · Staff engineer · Glasgow subscription app
7 / 10 — internal tool review
Warehouse-Native Measurement stopped the weekly argument with finance about timezone offsets. Dry delivery. Accurate.
Anonymous client in insurance
Funnel Cartography made us draw the age-gate. Conversion looked worse and more honest. Our CMO was not grateful, which I count as success.
Tom · Brighton
Session Replay Literacy was two weeks and that was enough. I left less tempted to narrate a user’s character from a truncated gesture. The UK masking examples were the useful part; the ethics lecture repeated a point I had already accepted.
Marta S., research ops, Cambridge
I took Observation Seat recordings while on parental leave. Naming law week is the one I replayed. I have not sat the live atelier yet; I am not sure I want the public critique, which is probably why I need it.
First name withheld · Birmingham
Case study · marketplace iOS
A Leeds marketplace arrived with four events that all claimed to measure “engagement” and none that marked a successful listing. During Event Taxonomy Atelier they dual-wrote a listing_published event with a closed enumeration for category, then deleted the paint-click events on a dated notice.
Six weeks after the board paper, weekly activation reporting shrank from fourteen charts to six. They still disagree internally about whether a draft listing should count; the atelier did not settle product politics, only the vocabulary.
Case study · grocery delivery
A Manchester grocery app used Amplitude for product and BigQuery for finance. Retention Geometry plus Warehouse-Native Measurement did not merge the stacks. It produced a one-page legend: which identity key each room used, and why seven-day retention in the console excluded a bot filter the warehouse applied.
The mild reservation from their analytics lead: the workshop assumed a SQL-comfortable partner in the room. When that person was on leave, the legend gathered dust until they returned.