“The activation lesson in Funnel Instrumentation finally gave us language for people who never hit the paywall. We still argue about Mixpanel vs PostHog in Slack; the class did not settle that.”
Priya N., growth lead, B2B design tool
Voices
What people said after they took the work back to their product
These are not a chorus. Length and tone differ because the weeks differed. One note below is mixed on purpose.
I came for Quiet Cohort Retention because our free users looked “fine” in a 30-day chart and then vanished from the upgrade path. The homework forced me to split “opened the app” from “did the job.” Leeds, rainy Tuesday, still using the worksheet.
Marcus, Leeds
★★★★☆
Night class, autumn table. The paywall timing studio was the week I needed. I marked it down a star because the privacy guest hour assumed more UK GDPR fluency than our US contractor had. The rest of the room seemed fine.
Platform review · learner in productivity software
Client in industrial IoT (free viewer seats, paid historian). We used the event-naming module to stop treating “dashboard loaded” as activation. No poetry here: our upgrade rate did not leap; our arguments in sprint planning got shorter.
Anonymous, industrial software
Loose leaf was the right size. I did not want a cohort. The PDF on unsigned sessions is the page I keep reprinting for new hires.
Elena Vázquez, staff designer who also owns analytics
Case: a notes app that counted the wrong first document
A twelve-person company in Bristol shipped a generous free plan. Their North Star was “documents created.” After Funnel Instrumentation they rebuilt the first-week story around “document shared with a second human,” which almost nobody on the free plan did. The paid plan’s collaboration tools were hidden behind a gate they had never instrumented as a decline.
They did not “fix conversion” in six weeks. They did stop reporting a cheerful activation rate that excluded the only behaviour that predicted invoices. The write-up from that table is now a sample in the flagship, with names removed.
Case: feature gates that fired on page load
A European developer-tools team (UK entity, remote) came to Side-by-side with 140 events and a paywall that evaluated on every route change. We spent the first morning deleting events that described CSS. The second day was a single experiment: delay the gate until a compile succeeded.
Their reservation, which they asked us to print: the clinic did not include a data engineer, so warehouse work landed back on their plate. Fair. We teach reading and naming; we do not become your dbt repo.