Voices
These notes refer to programmes actually taught in this house. Length and tone vary because people do. One of them is cooler than the rest on purpose.
★★★★☆
The session-window work in Mobile Session Reconstruction finally matched how our riders actually use the app between jobs. We had been using a thirty-minute timeout that glued two lunches together. After the programme we published the window in the taxonomy and the ops meeting stopped arguing about “drop-off” that was only a tea break.
Platform review · 4/5 · collected after the March room
Quiet Experiments was the first time anyone told me that an under-powered pricing test should be left unpublished. We still ran a smaller follow-up, but we stopped dressing a noisy conversion spike as a win in the all-hands. I wish the SQL excerpts had come earlier in the weeks.
Priya N., staff analyst, Bristol
Cohort Literacy for Product Leads gave me language for the board pack. I now write the sample size under the chart. Our CEO asked why we had never done that. I did not have a graceful answer.
Tom, Edinburgh
We came from grocery ops, expecting widgets. What we got was a slow argument about whether basket_updated meant a human or a sync job. I left annoyed at the pace and grateful for the memo. If you cannot spare a Friday in week three, wait for a quieter sprint.
Anonymous client in grocery delivery
Event Taxonomy Craft is short and slightly severe. We retired forty-one names. The room laughed at user_did_stuff, which we deserved. I would not send a junior alone; you need someone who can change the SDK afterwards.
A client in children’s education apps
Case notes
Helena’s team arrived with a funnel that showed trial starts twice — once at the paywall view, again at receipt validation. The Studio’s identity week forced a single trial_started with a source property. Week-four retention looked worse for a month, then became something they could defend. They did not “improve engagement.” They stopped lying to themselves about volume.
The grocery client (who asked to stay unnamed) had sessionization that treated a household iPad as one loyal shopper. Funnel Diagnostics Intensive spent four days on device graphs and household identity. The limitation they still print: they cannot yet separate teenagers adding snacks from the adult who pays. That sentence in the board pack was, they said, more useful than the old green arrows.