Consent mode gets installed the way most tracking does: someone ticks a box in the cookie banner plugin, sees the green “active” badge in Google Tag Manager, and moves on. A month later nobody can explain why Google Ads conversions dropped 30% in Europe and Canada while sales held steady. The two events are the same event.
What consent mode actually does
When a visitor declines cookies, consent mode tells Google’s tags to behave differently. In the basic setup, the tags don’t fire at all until consent is given. In the advanced setup, they fire but send cookieless pings with no identifiers, and Google uses those pings to model the conversions it can no longer observe directly. The modeled numbers show up in your reports as if they were real, with a small note you’ll never see unless you look for it.
Which of those two setups you have decides how much data you keep. Most sites I open are on the basic one by default, which means every declined banner is a visitor who never existed as far as Google Ads is concerned. Smart bidding then learns from the people who clicked accept, who are not a random sample of your customers.
The three places it breaks
The banner and the tags disagree. The consent management platform updates its own storage, but the signal never reaches GTM, or reaches it after the tags have already fired. Open the GTM preview, decline the banner, and watch whether the consent state actually changes before the first pageview tag runs. Half the time it doesn’t.
Meta and the others are left out. Consent mode is a Google framework. Your Meta pixel, TikTok pixel and server-side feeds need their own consent logic, and the plugin that handles Google’s usually doesn’t handle theirs. The result is a site that is compliant for Google and quietly firing everything else regardless of what the visitor chose.
Server-side forwards without the signal. If you run a server container, the consent state has to travel with each event to the server and be respected there. Forwarding declined-consent data to Meta from the server is not a clever workaround, it’s the same compliance problem with an extra hop.
How I check it
Three visits, with the debugger open. One accepting everything, one declining everything, one closing the banner without choosing. For each, I note which tags fired, what consent parameters they carried, and what landed in GA4’s DebugView and Meta’s test events. Then I compare the accept rate on the banner against the gap between platform conversions and real orders. If 40% of visitors decline and conversions are off by 40%, you’re on basic mode and losing all of it. If conversions are off by 10%, modeling is doing its job.
What it’s costing
The number that matters is not compliance, which you need either way. It’s the share of real customers your bidding never sees. On a site where a third of visitors decline, basic consent mode means smart bidding is optimizing on two-thirds of the truth, and the missing third skews toward privacy-conscious, higher-income segments in exactly the markets where consent banners are mandatory. I’ve written before about how attribution gaps compound. Consent is the one most teams don’t know they have, because the badge in GTM is green.
Working through something similar
This essay is one part of the measurement-first method. I’m always glad to compare notes on measurement and paid media. Get in touch → or find me on LinkedIn.