Reviews
61 written reviews sit on file. The average we quote is 4.3 out of 5, which is not a marketing round number. A few of the sharper ones are below.
I came for Funnel Grammar expecting a template pack. What I got was a two-hour argument about whether “viewed_paywall” and “paywall_seen” were the same event. They were not. Our conversion chart had been mixing them for eleven months.
★★★★★
Retention Atlas is the first course that made me plot a curve for people who completed onboarding on a Tuesday versus a Saturday. The Saturday cohort in our weather app behaved like a different product. We stopped blending them in the monthly pack.
Client in consumer meteorology, remote from Glasgow
Tom from Bristol. Sat Cohort Signal in the winter intake. The spoken readout practice was more useful than the slides. I still dislike presenting, but at least the sentences are shorter now.
House Commission found three events that were firing from a deprecated Android module. That was worth the fee. What it did not do — and the studio was clear about this — is rebuild our warehouse models. We had hoped they might stretch. They did not, and we had to hire that separately. Fair, if you read the scope twice.
Anonymous · fintech operations, London
Case notes
A council-facing app treated first-week open rate as health. After Cohort Signal, the team defined activation as successfully retrieving a collection schedule for the signed-in address. The seven-day open rate barely moved. The share of new accounts that reached that schedule event rose from 31 percent to 44 percent over a quarter, mostly because onboarding stopped asking for a marketing preference before the address.
Amelia Hough, who later left a public review, said the useful artefact was the exclusion rule: staff test accounts had been sitting inside the “new resident” cohort. Removing eighteen internal devices changed the curve more than any copy experiment that year.
Funnel Grammar students from a podcast company discovered that iOS used paywall_seen and Android used viewed_paywall, with different timestamps. Combined, the funnel looked healthy. Split by platform, Android appeared to convert from a ghost step. Engineering renamed both to paywall_presented with a platform property. Conversion did not improve; the argument did. Priya’s review on this page is from that circle.
Limitation they still live with: background audio interruptions were never instrumented, so a “session end” remains a blunt object. The studio logged that as out of scope for the clinic, which some participants found abrupt and others found honest.