Notes

What people said after the room went quiet.

We ask for specifics: a module, a fight, a metric they retired. Glowing blurbs without a product attached do not make this page.

★★★★★

“Event Taxonomy Lab forced us to kill ‘cta_click’ as a family of twelve near-duplicates. The argument lasted two stand-ups. The dictionary has lasted nine months.”

Maya H., analytics lead — Manchester

Sat Product Signal Studio in the autumn cohort. The activation week was the one that stayed with me: we had been celebrating first-session length while week-two return sagged. Amira made us write the reservation before the chart.

Tom · Cardiff

★★★★☆

Board Pack Analytics is short and slightly severe. I wanted more time on design of the slides themselves. What I did get — permission to put uncertainty on slide two — was the thing my CFO actually noticed.

Platform review · Workshop Seat, March
“Retention Cartography is not a visualisation class. It is an etiquette class for people who have been waving a single percentage at designers.”
Client in public-interest media

Case file 01

Product lead presenting to colleagues

A payments app that retired DAU

A London fintech sent three people to Product Signal Studio after a board asked for “one number to watch.” They arrived with DAU printed in 72-point type. By week two they had a working definition of funded-and-returned, which was uglier and more predictive.

The reservation they kept: the new metric would look worse for a quarter while dormant wallets were no longer counted as health. The CFO signed the board pack anyway. We are not allowed to name the app; the method is in the syllabus.

— Narrated by Amira Ellison, with the trio’s permission to describe the metric change only.

Case file 02

Team working at a long table with laptops

A health app and a funnel that lied by omission

During Attribution without Theatre, a growth lead from a United Kingdom health app brought a five-step funnel that began after onboarding consent. Completions looked heroic. The missing step was the consent screen itself, where a third of people left and never became a row in Mixpanel.

They did not “fix conversion.” They redrew the funnel to include the refusal, then wrote a board sentence about trust rather than about ads. The work was slower than a media-buying tweak. It was also the first time engineering and growth shared a diagram.