Voices
What people kept after the recordings expired.
Consent Signal Architecture’s warehouse-contract module is the only class I have taken that made finance stop asking why “sessions” collapsed on Monday. We had labelled declined traffic as a tracking outage.
Jo, Glasgow. The CMP-to-warehouse mapping lab was slower than I wanted. Worth the drag: we found duplicate grants when the banner retried.
★★★★☆
Solid on fallback events. I still wish the open cohort spent a session on consent mode myths instead of sending us to the journal. Fine, but I paid for live time.
Client in insurance. We brought a 90-event dictionary to the monthly clinic. Half of it fired before the prompt. The punch list was uncomfortable and accurate.
Case: publisher homepage after a banner redesign
A UK newsroom cut “engaged sessions” by a third after a clearer reject control. Leadership treated it as a product failure. After Field Notes plus a private taxonomy pass, they split the dashboard into coverage-labelled panels: essential logs, granted analytics, and a reject-majority strip fed by first-party, non-identifying counts. Advertising ops still wanted the old number; the case did not convert them. It did stop the product team rewriting the homepage to chase a metric that had been consent theatre.
Case: retailer experiment engine
A grocery group’s A/B platform allocated variants before the CMP returned. Experimentation under consent did not “fix” the vendor. The team added a hold-out that only entered analysis after a consent snapshot existed, and they published a sample-ratio warning on every test brief. Conversion lift claims shrank. The reservation they still hold: native app traffic remains outside that contract.