12 May 2026

Signal loss after the cookie prompt

Abstract data visualisation on a large display

A product team in Birmingham brought us a funnel that “died” between landing and article scroll. The overlay sat on the first paint. Analytics tags waited for a grant. Everyone who tapped Reject vanished from the chart, then reappeared in finance’s ad invoices as if they had never arrived.

That is not drop-off in the interaction sense. It is a consent cliff mislabelled as a UX failure. Consent-Aware Analytics Design starts by drawing two timelines: what the person saw, and what the tag manager was allowed to hear.

Compare server access logs (no identifier beyond what you already need to serve the page) with client hits that carry a consent snapshot. If the server still shows a healthy scroll depth proxy — such as requests for later image breakpoints — while analytics reports a ghost town, you are looking at coverage, not boredom.

The repair is unglamorous. Label the dashboard. Stop asking design to “fix bounce” that was never a bounce. If you need a product metric for reject-majority traffic, design a fallback event that does not impersonate a cookie ID. We practise that split in Consent Signal Architecture.

A mild warning: access logs are easy to over-read. A 200 on a CSS file is not an engaged minute. Keep fallback signals coarse, or you will rebuild surveillance with extra steps.