Kettu is a services marketplace connecting households with verified technicians — plumbing, appliance repair, electrical work. Supply was healthy and paid acquisition was bringing people into the app, but the share of new signups who ever booked a first job had been falling for three quarters while the team kept buying more traffic.
The funnel was leaking somewhere between account creation and first booking and nobody could say where, because the events had been instrumented by three different teams over four years and no two of them agreed on what counted as a signup. Lifecycle email had been switched off entirely after a deliverability incident and never turned back on.
I rebuilt the event taxonomy first, which meant an unglamorous month spent mapping every existing event onto a single definition and deprecating the rest. With numbers we could trust, the leak was obvious: users who did not book within 48 hours almost never came back, and nothing reached them inside that window. We rebuilt lifecycle on a warmed domain with authentication done properly, then replaced the generic welcome series with three messages tied to the category the user had actually browsed. Separately we moved address capture out of signup and into the booking flow, which removed the single largest drop-off step in the funnel.
Signup-to-booking conversion went from 2.1% to 5.4%. Email open rate recovered from 11% to 34% once the domain was repaired and the messages were relevant to what people had looked at. Monthly signups grew from 18,000 to 41,000, though the conversion change mattered considerably more than the volume did.
Most of the value came from deleting things — events, form fields, an entire welcome series. The instrumentation month produced no visible output and was the only reason any of the rest could be diagnosed, which is a difficult thing to put on a roadmap.