1. The one difference that matters
Google's GA4 help documentation draws the line clearly. In an Advanced implementation, Google tags load when the page loads, before the consent dialog is answered, and when consent is declined they send cookieless pings instead of cookies. In a Basic implementation, Google tags stay blocked until the visitor grants consent, so nothing is sent beforehand. Everything else, including what GA4 can model, follows from that one timing choice. The consent signals themselves are the ones set up in the privacy-first analytics blueprint.
2. What each mode lets GA4 model
With Advanced, GA4 can use behavioral modeling to fill gaps left by users who declined analytics consent, estimating their activity from the patterns of those who accepted, and it can also model conversions. With Basic, modeling is limited to conversions, and only where some conversion data was collected before consent was denied. So Advanced gives you more modeled data to work with, and Basic gives you less while sending nothing before a choice is made.
3. The part nobody can audit
Google's documentation does not publish a minimum event or user count for modeling to work, and the modeled numbers are Google's estimate, produced by a process you cannot inspect. I find that the right posture is to treat modeled data as a reasonable estimate that fills a gap, never as a measurement you can reproduce. If a modeled figure drives a budget decision, say so in the readout. You will also see confident claims online about how much lost data modeling recovers; I would not repeat any of them without a primary source, because the vendor itself does not state one.
4. Choosing is a policy decision first
Advanced means sending pings to Google before a visitor has made a choice. Whether that is acceptable depends on your jurisdiction, your legal team's reading of the rules, and your own risk tolerance, not on what the dashboard looks like. I have seen this decision made purely on measurement grounds and then reversed once legal weighed in, which cost weeks of instrumentation rework. Settle it with legal and privacy up front. If you serve visitors under US opt-out signals as well, remember that Global Privacy Control is a separate mechanism from the banner choice.
5. Verify what your setup actually does
Do not assume which mode you are in. Open your site with a tag debugging tool, decline consent, and watch whether any Google requests fire before you choose. Compare that against what you intended, and re-test after every banner or tag change, since a small edit can silently move you from one mode to the other. The consent testing playbook covers the instrumentation side, and the server-side GTM primer covers what changes if tags move off the browser. My take: pick the mode your legal team will defend in writing, then measure around its limits instead of fighting them. A limitation of this article: it describes Google's documented behavior as I read it, and the details can change, so recheck the current help page before you commit.
How different teams plug in
Consent mode touches several owners, and the choice sticks better when they agree on it together:
- Legal and privacy decide whether sending pings before consent is acceptable and document the reasoning.
- Engineering implements the consent defaults and confirms nothing fires earlier than intended.
- Analytics labels modeled numbers as estimates and explains the gap to stakeholders.
- Marketing understands that campaign reporting now includes modeled data and plans around it.