Reviews

After the ledger, not during the applause.

These notes refer to programmes on this site. We keep the mild complaints in public; a studio that only prints glow looks like a brochure.

The week on identity graphs moved faster than I could keep notes, and I still reach for the workbook more than I would like. Even so, the cohort-window exercises changed how our iOS team reports month-one retention.

Helen Cartwright, Head of Product, Bristol fintech

★★★★★

Funnel Forensics finally explained why our “checkout started” count was larger than “item viewed”. Duplicate event names. Three stars for the homework load, five for the diagnosis.

4.4 average from that cohort’s written slips

I came for Mixpanel tricks and was refused them until I could define a session without pointing at a vendor toggle. Irritating in week two. Correct.

James, Edinburgh

Our crash-free session rate looked elite because we were dropping background wakes from the denominator. Session Health Desk made that visible. We did not enjoy the meeting that followed.

Client in consumer health, Midlands

Case · grocery

A 28-day curve that only existed in January.

A Manchester grocery app joined Instrument Bench after a board pack showed “record retention”. The ledger exercise split cohorts by store-listing source and by calendar week. The January gym-adjacent recipe feature had pulled a population that never intended to grocery-shop weekly. By March the curve collapsed to its old shape.

They did not need a new North Star. They needed a window that did not treat a seasonal tourist as a resident. The team now publishes two retention families: residents and listing visitors. Marketing still argues. The argument uses the same clock.

Team discussing work around a table — Paired seats: product and engineering in the same critique.

Two colleagues reviewing a document — Close-out letter after a Kernel Residency.

Case · rail companion

Store-to-open looked healthy because crashes happened after the first frame.

A regional rail app measured “successful open” as process start. Crash-free sessions, counted later, told a harsher story on older Android devices. The residency rewrote the open event to mean first interactive frame and moved crash-free onto the weekly ledger beside punctuality complaints.

Limitation they still live with: the studio did not rebuild their media attribution. Paid install quality remains a separate, messier book.