Voices

What people kept after the last session.

These notes refer to programmes on this site. They are not a wall of five-star paste. One or two still argue with us, which is how a measurement house should sound.

“Identity & Sessions was the module I expected to skim. It is the one I now make every new analyst read before they touch our app.”

Amelia Hart · Head of Product, coastal banking app

I still think the homework load in week four of Signal Architecture is slightly proud of itself. The funnel reconstruction that came out of it, however, is the only one our board has ever understood.

Jonah, Manchester
★★★★☆ · 8/10 alumni board

Clear on privacy labels. Less romantic about Mixpanel than I wanted, which I later needed.

Posted as “rowan.k” on the closed alumni ledger, March 2026

A client in subscription fitness asked us to stop reporting a blended retention number. After Retention as Narrative, we finally had language for why that number was flattering the wrong week.

Anonymous client in consumer fitness

The House Taxonomy page is now pinned in our engineering wiki. I did not expect a teaching site to produce a document we would actually ship against.

Priya, Leeds

Case notes

Two rooms, after the course

Small product team talking around laptops in a bright studio

A regional rail app and a funnel that lied politely

The team arrived with a checkout funnel that “converted” at a handsome rate because it ignored ticket-type switches mid-flow. During Funnel Reconstruction they rebuilt the path around the actual states a passenger occupies. Conversion fell on paper and rose in the operations meeting, which is the correct direction for an honest instrument.

They did not replace their vendor. They replaced the story the vendor was being asked to tell.

A catalogue retailer that had instrumented its own confusion

Two hundred and nineteen events, many of them button-level, none of them owned. The Conservatory desk spent three studios deleting. What remained was a house taxonomy of thirty-one events and a weekly reading of two cohort charts. Staff still argue about names; they no longer argue about whether a chart exists.

The mild cost: the first month after deletion felt like flying without instruments. That feeling passed when the remaining signals started matching what shop-floor staff already knew.

Team discussion at a conference table with notebooks open