Why conversions vanish after a cookie banner
Why GA4 and ad platforms lose conversions once a consent banner goes live, and how MIRA FIVE counts visits without cookies and people only with consent.
With consent means a visitor has agreed to be recognised, so their visits can be joined into one person with a history across days.
Once a visitor agrees, an analytics tool may store an identifier in their browser and recognise them when they come back. That is what turns visits into people. Return visits, the path from first visit to purchase and funnels that span several days all depend on it.
Consent is usually asked for in a banner or a settings panel. The share of visitors who agree varies widely by audience and by how the question is asked, so a good tool tells you how much of its data rests on consent.
Visitors who consent are followed as people across visits. On the site, that means full mode: switch the website source to Full under Data → Sources, then set data-mode="full" on the script tag, or mode: 'full' in an SDK. Nothing is sent or stored until the visitor answers, and your banner passes the answer with mirafive('consent', …): true for statistics only, false, or { statistics, experiments, targeting } scope by scope. Ids live in localStorage, never in cookies. When the person signs in, mirafive('identify', userId) links their history to their account, and reset() on sign-out forgets it.
Visits without consent are counted in visits, pages, channels and events where a page runs the default setup, as described under consentless analytics. Counting visitors without consent while following those who agree takes a website source set to Full, with each page choosing the mode per visitor; a data-mode="full" tag alone sends nothing for visitors who decline. The dashboard then shows the share of traffic with consent, for example “85% with consent”. A/B tests in the browser count only people who consented to statistics and experiments.
Emil Becker’s page in the Acme Shop shows the first visit through an Instagram ad, the newsletter click and the return through Google search before the €328 order. Joining those visits into one person needed Emil’s consent.
Why GA4 and ad platforms lose conversions once a consent banner goes live, and how MIRA FIVE counts visits without cookies and people only with consent.
How long an A/B test should run, planned before it starts, in whole weeks, with the smallest change your traffic can detect worked out for 2, 4 and 8 weeks.