You open Tag Manager preview. The tag fires. The trigger looks right. Everything is green. Then the client asks why conversions are down thirty percent. You have nothing to show them. On paper, nothing is broken.
This is a common gap in Consent Mode v2 setups. It will not show up in preview mode at all.
A fired tag does not mean Google actually used the data. Consent Mode v2 checks four separate signals before ad data counts as usable:
analytics_storage, controls whether analytics cookies can be used
ad_storage, controls whether advertising cookies can be used
ad_user_data, controls whether user data can be sent to Google for ads
ad_personalization, controls whether that data can be used to personalize ads
If any of these were denied, or never explicitly set, the tag fires fine. It just contributes nothing usable to remarketing or attribution.
The mistake that causes this most often
The consent update triggers on page load instead of the actual button click. That creates a race condition. The tracking pixel fires a split second before the CMP registers the real choice. Preview mode does not catch this, because it does not simulate that timing gap. Fix: trigger marketing tags off the consent update event specifically, never off page view.
What changed in June 2026
Before June 2026, a GA4 setting called Google Signals could still pull in some ad data, even with an incomplete consent setup. That workaround is gone now. ad_storage alone decides whether advertising data reaches Google Ads. If your dataLayer wiring is off, nothing catches that gap anymore.
Worth checking beyond the code
GDPR also requires reject to carry equal visual weight to accept. Not smaller text, not a duller color. Several EU fines were issued specifically over banner design, not just tracking behavior. Worth flagging to whoever owns the banner, even if it is outside your usual scope.
Once the signal is wired correctly, Google's modeling recovers a real share of what was lost. A median lift of around 17 percent in reported conversions has been cited once modeling activates properly. That only works if the underlying signal was accurate to begin with.
If maintaining three separate consent-to-signal mappings by hand is not worth your time, Seers handles that pipeline directly, mapped to Google, Meta, and Amazon. There is a free scan that flags scripts firing outside the consent event, worth running against your current setup either way.
Non-technical breakdown of the fines tied to this exact issue is here.
Top comments (0)