GA4 sessions fell the week a consent banner went live. That timing is a clue, not a diagnosis. Visitors may still be arriving but no longer measurable in the same way; a tag may be misconfigured; demand or campaigns may also have changed. The first job is to separate those possibilities without changing the consent choice merely to make a chart recover.
Google distinguishes basic and advanced consent mode. In a basic implementation, Google tags are blocked until consent, and a visitor who declines sends no data to Google through those tags. In an advanced implementation, tags load with a default state and can send measurements without cookies while consent is denied. Which implementation is appropriate is a company and legal decision; this article cannot decide the applicable law or certify compliance.
Build a dated evidence ledger
Record the banner release time, affected pages, tag/container changes, campaign changes, and the date range of the GA4 report. Compare like-for-like weekdays and channels, but do not assume the same consent rate or visitor mix. Use three hypotheses:
| Hypothesis | What would support it | What would weaken it |
|---|---|---|
| Real visitor demand fell | Independent channel evidence and qualified inquiries also decline, with campaigns, seasonality, and site availability checked. | Only GA4 changes while other relevant signals stay broadly similar. |
| Measurement coverage changed as expected | The implementation sends less data for visitors who decline; GA4 loss begins with the consent release. | The drop also appears among test visits that granted consent. |
| Consent or tag wiring broke | A permitted test action is missing, duplicated, or attributed incorrectly after consent is granted. | The same test action and event appear correctly in each approved state. |
Qualified inquiries from a form or CRM are a useful independent outcome count, not a substitute for visits: stable leads cannot prove stable traffic, and a falling lead count may reflect either fewer visitors or a broken form. A Search Console click trend covers only Google Search, not every acquisition channel. Record each source's scope next to its number so a partial metric is not mistaken for the whole site.
Test both consent choices on the actual buyer path
Use a clean browser session and a synthetic inquiry. First decline analytics consent, then repeat with consent granted. On each path, record the banner state, page, event name, form delivery result, and whether a conversion appears in the relevant debugging tools. Google's Tag Assistant verification guide describes how to inspect the default and updated consent states. Its implementation guide emphasizes setting a default state before measurement and updating the state on the page where the visitor acts. A late update or one lost during navigation can make reports incomplete.
Do not expect every denied-consent visit to reappear through modeling. Google Analytics' behavioral-modeling help lists eligibility conditions including advanced consent mode across pages, at least 1,000 daily events with analytics_storage='denied' for seven days, and at least 1,000 daily users sending granted-consent events for seven of the previous 28 days. Even meeting those published prerequisites does not guarantee modeled data. Many smaller sites cannot use modeled totals to reconstruct their missing visits.
Decide what the report can still answer
If permitted test events fail, repair the tag or consent integration and annotate the change. If the implementation behaves as approved but fewer visits are observable, report the coverage break alongside other evidence; do not splice pre-banner and post-banner sessions into a smooth growth chart. If independent demand evidence also falls, investigate that separately. Re-run the same test after any banner, tag-manager, checkout, or form change.
The qualified inquiry baseline guide shows why lead quality must be tracked beyond pageviews. The contact-form delivery test checks whether the business received an inquiry at all. This article explains why a GA4 drop after consent may be a measurement problem, a real visitor problem, or both.
Keep the release date, tag setup, one synthetic test result, and the report range beside the chart. Those four pieces let the next analyst repeat the diagnosis. For ARCHV website work, analytics and consent behavior should be written into the actual proposal and tested against the approved policy.