How consent mode works
Consent mode is a signalling layer sitting between a cookie banner and Google’s tags. Rather than the banner simply blocking or allowing scripts, the site declares a consent state that every Google tag reads before it does anything. That state covers a small set of named permissions: storage for analytics, storage for advertising, and separate signals for sending user data and for personalisation.
Two things have to happen, in order. A default state is declared before any tag loads, saying what is permitted while the visitor has not chosen yet. Then, the moment the visitor accepts or refuses, the banner sends an update and the tags change behaviour on the spot. In the stricter setup nothing fires until consent arrives; in the looser one tags load in a restricted mode, sending pings that carry no identifiers, and Google uses those to model what it cannot observe.
Why consent mode matters
If you advertise to people in the European Economic Area or the UK, Google requires these signals before it will keep offering its measurement and audience features for that traffic. Without them, remarketing lists and conversion measurement for those visitors degrade rather than quietly carrying on as before.
It matters outside Europe too, in a quieter way. A business in Kathmandu selling to clients in the UK, Australia or the EU inherits the same obligations for those visitors even though the rules at home are different. Setting it up once is far cheaper than retrofitting it after a client’s legal team asks the question, or after Google limits what the ad account is allowed to use.
Common mistakes with consent mode
The most frequent is installing a banner and assuming the work is finished. Plenty of banners look compliant while sending no consent signals at all, or sending them after the tags have already fired, which amounts to the same thing. If the default state is not declared before the first tag loads, the sequence is broken however polished the banner looks.
The second is never testing the refusal path. Teams accept the banner, watch data flow in and call it working, when the important question is what happens when somebody declines. The third is treating modelled figures as the same kind of number as observed ones. Modelling is an estimate Google fills in where consent was withheld, and it belongs in a report with that label attached to it.
How to get it right
Decide first which visitors are genuinely in scope, because that answer shapes everything else — a banner shown to every visitor in every country is a conversion cost you may not need to pay. Then implement the default and update calls through Google Tag Manager rather than scattering them across page templates, so there is one place to inspect and one place to change.
Test both paths deliberately. Accept, and confirm the tags fire with storage available; refuse, and confirm they fall back to restricted behaviour instead of either firing normally or disappearing altogether. Expect observed conversions to fall afterwards, and warn whoever reads the reports before it happens, because an unexplained dip gets blamed on the campaigns. Keep a written record of what was configured and when, filed with your conversion tracking notes.