★★★★★★★★☆☆ 8/10
“Signal to Send’s incrementality module stopped us from celebrating a win-back that was stealing from same-day receipts. We still argue about sample size, but at least we argue about the same definition.”
Reviews
Notes from catalogues, not from a prompt.
Attribution styles differ because people share differently. Where a company asked to stay unnamed, we kept the industry and city.
“The Delivery Funnel workshop was only three days and still the first time our ESP ‘delivered’ column got split from lock-screen render. I work in Chiang Mai; the ICT examples did not feel Bangkok-only.”
Waranya Prasert, CRM lead at a Bangkok grocery app, stayed after module 2 to rebuild Friday lunch blasts. “We had been reporting a healthy tap rate while render was quietly falling on older Android skins. The worksheet made that visible in one afternoon. I would not call the lab gentle — Nalinee marks incomplete funnels — but the Friday meeting got shorter.”
“Quiet Hours, Loud Data helped. The reservation: our ESP still cannot honour provincial time zones, so the studio became a briefing document for engineering rather than a toggle we could flip. That was disappointing, and also the honest outcome.”
Case study 01
Lunch blasts that looked busy and moved little
A Bangkok grocery catalogue sent a weekday 11:40 push for twelve months. Open rate sat in a comfortable band. After Signal to Send module 2, the team split accepted / delivered / rendered and found older Android devices were absorbing “delivered” without ever showing the card.
They did not rewrite copy first. They cleaned tokens older than 120 days, moved the remaining send to 11:05 for suburban ICT users, and only then tested a shorter title. Basket influence on the following day rose enough that finance agreed to keep the Friday slot — with a render floor written into the weekly report.
Shared with permission. Company name withheld.
Case study 02
Token cliff after an iOS certificate rotation
A travel super-app watched delivery fall nine points in a week and assumed creative fatigue. Module 4 of Signal to Send — token health — matched the drop to an APNs certificate rotation that left a slice of devices unregistered.
The CRM team had no backend seat. They borrowed an engineer for two afternoons, exactly the limitation several alumni mention. Once invalid tokens were purged, copy tests became readable again. Mee, product operations in Phuket, later wrote that the lab “did not fix the certificate, it stopped us from rewriting headlines for a plumbing problem.”
