Consent gap
The visit happened. The cookie said no.
GA4 waits for consent and then for a tag that blockers already killed. The HTML still ran. Cookie Free Analytics records that fact.
Where the row disappears
A hundred people hit /pricing. Forty click Reject. Ten run an extension that swallows googletagmanager.com. A few use Safari’s defaults. GA reports a number that looks precise and is a sample.
Conversion rate goes up because the denominator shrank. A privacy newsletter looks like a bad channel. You optimize the people who click Accept.
What this does to a campaign
The paid click still landed. The banner asked. They said no. Google Ads and Meta did not get a conversion ping. GA did not store the UTM. In those tools the campaign looks like a miss, and return on ad spend is the Accept pile.
Consent mode can fill some of that hole with a model. A model is not a count of the people who opened the page. Cookie Free Analytics records the load and the five UTM tags on the landing. If you fire cfa('signup') without waiting for a cookie, that goal is in the same count. We do not ingest ad spend or revenue, so we do not print ROAS. We print the visits cookie reports lost — the ones you still paid for.
We will not tell you to delete a banner that still covers Google Ads. Keep those tags if you buy ads. Keep us as the count of what actually loaded.
This is not a GA setting
Consent mode can model some of the hole. Modeling is not a count. First-party cookies still need a yes in the EU if they identify someone. The structural problem is a tool that needs a cookie to begin.
We do not need that cookie. The script fires with the page. You can leave GA in place for ads. Use us as the baseline: what actually loaded. The counts live in the EU, on European infrastructure — not a US property with consent mode papering over the hole.
How to see your own gap
- Keep GA.
- Paste our script.
- Wait a week.
- Compare pageviews. The difference is mostly reject + block, not “bots we invented.”
We will not invent a 400,000-pageview case study we did not run. Your logs are the case study.