AdPix Blog

Cookie consent analytics: what your banner does to your numbers

Cookie consent analytics: what a banner removes from your reports, basic vs advanced consent mode, GA4 modeling limits and how to correct for consent rate.

Updated 10 min read

Key takeaways

  • A consent banner does not shrink your data evenly. Channels, devices and visitor types accept at different rates, so the share of conversions each channel gets in analytics shifts, not just the totals.
  • In Google's basic consent mode, a visitor who declines sends Google nothing, not even the consent choice. In advanced mode, Google tags still send cookieless pings while consent is denied.
  • GA4 only models unconsented behavior when the property uses the advanced implementation, has at least 1,000 events a day with analytics_storage denied for 7 days, and has at least 1,000 consented daily users on 7 of the previous 28 days.
  • Consent rate is consented visits divided by all visits that saw the banner. Measure it per channel and device, and use your order backend, which records every paid order, as the check.
  • Correcting observed numbers with one site-wide consent rate fixes nothing about channel mix. Dividing each channel by its own consent rate gets much closer, as long as consent and buying are not linked within that channel.

Cookie consent analytics is the work of knowing what your consent banner removes from your reports and correcting for it. When a visitor declines analytics cookies, their visit either vanishes or arrives as an anonymous, cookieless hit, depending on how your tags are set up. The damage is uneven: acceptance differs by channel, device and visitor type, so a banner changes which channel looks profitable, not only how big your totals are. Measure your consent rate per segment, compare against your order backend, and you can put most of the picture back together.

Consent rate is the share of visits, among those that were shown a consent banner, in which the visitor allowed analytics (or the relevant category). Everything you see in analytics reports about unconsented visitors is either missing, cookieless or modeled.

Cookie consent removes or degrades the visits of everyone who declines, and it does so at different rates for different groups. The visible effect is lower users and sessions. The less visible effects matter more for decisions:

  • Totals drop. Users, sessions, page views and conversion counts fall roughly in line with the share of visitors who decline.
  • Channel mix shifts. A paid social campaign landing mostly first-time mobile visitors may see far lower acceptance than returning visitors arriving directly. Analytics then gives that campaign a smaller share of conversions than it really drove.
  • Per-user metrics move. Without a cookie, a returning visitor cannot be recognized, so new-user share, sessions per user and retention numbers change meaning. Repeat-purchase figures such as customer lifetime value are safer to build from store orders than from analytics users.
  • Ad platform numbers drift away from analytics. Ad platforms count clicks; analytics counts consented sessions. The gap between them grows with every visitor who declines.

The last point is a common source of panic. If you already compare tools, the guide to GA4 data discrepancies lists consent next to session rules, bots and blocked tags as one of the causes you should expect. Bots are a separate problem that inflates numbers rather than shrinking them; the bot traffic detection guide covers that side.

Google’s consent mode lets you pass a visitor’s choice to Google tags, which then change what they store and send. Google’s consent mode overview lists seven consent types:

Consent type What it controls (per Google’s docs)
ad_storage Storage such as cookies related to advertising
analytics_storage Storage related to analytics, for example visit duration
ad_user_data Consent for sending user data to Google for online advertising
ad_personalization Consent for personalized advertising
functionality_storage Storage that supports site functions, such as language settings
personalization_storage Storage for personalization, such as video recommendations
security_storage Storage for security, such as authentication and fraud prevention

ad_user_data and ad_personalization are the two that Google added in November 2023, often called consent mode v2. Google’s analytics help says Analytics “expects the ad_user_data consent mode parameter” for ads measurement of users in the European Economic Area (EEA).

The bigger decision is how your tags behave before and after a choice. Google describes two implementations:

Basic consent mode Advanced consent mode
Before the visitor chooses Google tags do not load Tags load with defaults, usually denied
Visitor declines “No data is transferred to Google at all – not even the consent status” Tags “send measurements without cookies”
Visitor accepts Full measurement Full measurement
Modeling General model (less detailed) Advertiser-specific model (more detailed)
GA4 behavioral modeling Not eligible (requires tags to load in all cases) Eligible if the thresholds below are met

Basic mode is simpler and sends less. Advanced mode keeps a count of unconsented activity without cookies, which is what modeling and consent-rate measurement need. Which one you are allowed to use is a legal question for your market, not a reporting one.

Does GA4 fill in the data from visitors who decline?

GA4 can estimate unconsented behavior, but only for properties that meet Google’s thresholds, and many small and mid-sized sites never do. Google’s behavioral modeling page lists three prerequisites:

  1. Consent mode uses the advanced implementation: tags load before the consent dialog appears and “load in all cases, not only if the user consents.”
  2. The property collects at least 1,000 events per day with analytics_storage='denied' for at least 7 days.
  3. The property has at least 1,000 daily users sending events with analytics_storage='granted' for at least 7 of the previous 28 days.

Google adds that meeting the threshold for more than 7 days may be needed, and that the data may still not be enough. When modeling runs, you only see it if you choose the Blended reporting identity, and the help page notes features that do not use modeled data, including audiences, retention reports, predictive metrics and the BigQuery export. Modeling is not the same as sampling: modeling estimates visitors GA4 never observed, while sampling reads only part of the data it did collect; the GA4 data sampling guide covers when that happens and how to spot it.

So for a store with 600 consented visitors a day, GA4 shows observed data only. Everything below about measuring and correcting for consent rate is your job, not Google’s.

Consent rate is consented visits divided by all visits that were shown the banner, measured per segment rather than site-wide.

consent rate = visits with analytics consent granted / visits that were shown the banner

You need a count of the denominator, which analytics alone cannot give you under basic mode, because declined visits never arrive. Three practical sources:

  • The consent tool’s own log. A consent management platform (CMP) records each choice. Export accepts and declines per day, device and, if it can, landing page or campaign.
  • Cookieless pings in advanced mode. Unconsented hits still reach Google, so you can compare consented with total activity.
  • Your backend. Orders, sign-ups and leads are recorded by your shop or CRM whether or not a visitor accepted cookies. Observed conversions in analytics divided by backend conversions gives you a consent rate among buyers, which is the number that matters for revenue reports.

Keep a weekly table of consent rate by channel and device type. A sudden drop usually means a banner change, a new campaign audience or a broken update call, and it will look like a traffic drop in every other report.

Worked example: why one site-wide correction fails

Take a hypothetical online store. The numbers are invented to show the arithmetic. In one month it gets 20,000 visits from three channels and records 400 paid orders in its shop backend, with an average order value of 1.5 million toman. Its banner log gives a consent rate per channel.

Channel Real visits Consent rate Visits in analytics Real orders Orders in analytics
Instagram ads 10,000 45% 4,500 120 54
Google organic 6,000 70% 4,200 140 98
Direct and returning 4,000 80% 3,200 140 112
Total 20,000 59.5% 11,900 400 264

Inside each channel the conversion rate looks right, because acceptance here is assumed to be unrelated to buying: Instagram converts at 54 / 4,500 = 1.2% in analytics and 120 / 10,000 = 1.2% in reality. The channel mix is what breaks. Analytics credits Instagram with 54 / 264 = 20.5% of orders; the real share is 120 / 400 = 30%.

Now the money. Say Instagram spend was 60 million toman. Real Instagram revenue is 120 × 1.5M = 180M toman, a ROAS of 3.0. Analytics shows 54 × 1.5M = 81M toman, a ROAS of 1.35. A team with a break-even ROAS of 2 (see the ROAS formula guide) would cut a campaign that is in fact profitable.

Two ways to correct it:

  1. One site-wide rate. Divide observed orders by the overall consent rate: 264 / 0.595 ≈ 444 orders. The total is now 11% too high, and Instagram’s share is unchanged at 20.5%, so the budget decision is still wrong.
  2. Per-channel rates. Divide each channel by its own rate: 54 / 0.45 = 120, 98 / 0.70 = 140, 112 / 0.80 = 140. The sum is 400, which matches the backend, and each channel’s share is right.

The per-channel method rests on one assumption: within a channel, visitors who accept and visitors who decline buy at the same rate. Logged-in, loyal customers may accept more often than first-time visitors, which breaks it. The check is always the backend. If corrected orders sum to far more or less than the orders you actually shipped, your segments are too coarse, and splitting by device or new versus returning usually closes the gap. The same idea feeds attribution models: no model can credit touchpoints it never saw.

How do you set up a banner so the data stays usable?

A banner can be legally sound and still wreck your data through ordering mistakes. Google’s setup guide says the order of the consent code “is vital”, and most broken setups fail on these steps:

  1. Set defaults before any measurement. Call gtag('consent', 'default', ...) on every page before config or event commands. Out of order, Google says, “consent defaults won’t work.”
  2. Scope defaults by region when your rules differ by region. The region parameter takes ISO 3166-2 codes; a default without a region applies to everyone not covered by a more specific one. Google calls it best practice to scope defaults to the regions where you show a banner.
  3. Give an asynchronous CMP time. wait_for_update (for example 500 milliseconds) tells tags to wait for the update call before sending.
  4. Send the update as soon as the visitor chooses, and again whenever the choice changes, before the page unloads.
  5. Test both paths. Accept once and decline once in a clean browser, then confirm what arrives. In GA4, Admin → Data collection and modification → Consent settings shows whether ads and analytics consent signals are set up correctly.
  6. Log the consent rate weekly per channel and device, as above, and annotate banner changes in your reports.

The website tracking guide puts the accept-and-decline test into its full tracking QA checklist. Declined visitors still place orders, so keep the source of each order in your own database as well; the guide to first-party data shows which records survive consent choices and browser limits, and how to rebuild a channel report from them. Run the same accept-and-decline test on your custom events too: the event tracking guide shows how to test each event in a debug view before release, so a purchase or lead event that only fires after consent shows up before it skews your reports. Moving a tag to a server does not change any of this: the server-side tracking guide explains why an event that needed consent in the browser still needs it when your server sends it.

AdPix is a web analytics platform that replaces Google Analytics 4: realtime, acquisition, engagement, conversion and e-commerce reports, plus attribution, lift testing and marketing mix modelling, with no data sampling and a free plan. GA4 itself ships no banner; Google’s consent mode passes choices to Google tags, and you still need a CMP to ask.

AdPix includes a cookie consent banner on every plan, including Free, built into the tracker rather than added as a separate tool. According to the AdPix consent banner docs, you set four categories (Necessary, Preferences, Statistics, Marketing), write the banner text and choose opt-in or opt-out. Under opt-in the tracker fails closed: until a visitor chooses, nothing non-necessary runs and no page view is sent, so expect a real drop in counted traffic, exactly as in the example above. The docs also say region-based switching is not implemented yet, so a site with European visitors should use opt-in for everyone. Some features depend on consent too: AdPix’s browser-side invalid-traffic checks only run for visitors who granted the Marketing category. The AdPix cookie consent guide shows what each category gates in the tracker, what the consent log records and how to read its acceptance rates as a measure of what your reports are missing.

Because consent and analytics live in one product, the banner log and the reports describe the same visits, which makes per-channel consent rates easier to keep. The AdPix features overview shows where consent sits next to funnels, revenue reports and invalid-traffic detection, and AdPix pricing lists the plans. If you are adding the tracker for the first time, the AdPix install guide covers the snippet, GTM and the settings worth changing in the first hour. If your site serves Iran and still runs GA4 alone, read the guide to Google Analytics risks in Iran before you change anything else. For a wider comparison with GA4 and Matomo, including how each handles consent, read the Google Analytics alternative guide.

Start this week with the cheapest step: export last month’s orders from your shop backend, put them next to the orders your analytics recorded per channel, and divide. That ratio is your consent rate among buyers, and it tells you how much of every channel report you have been missing.

Measure your site with AdPix today

The free plan covers 100,000 page views a month for each website, no credit card. Setup takes under five minutes.

Frequently asked questions

How does cookie consent affect analytics?

Visitors who decline analytics cookies either disappear from reports or appear only as cookieless, unidentified hits, depending on how your tags are set up. Totals drop, and because acceptance differs by channel and device, the channel mix and attributed revenue in your reports also change.

What is the difference between basic and advanced consent mode?

In basic consent mode, Google tags are blocked until the visitor makes a choice, and nothing is sent to Google if they decline. In advanced consent mode, tags load with consent denied by default and send cookieless measurements, then send full data once the visitor grants consent.

What is a good cookie consent rate?

There is no reliable universal benchmark, because acceptance depends on the banner design, the audience, the device and the legal setup. Track your own consent rate per channel and device every week and treat sudden changes as a measurement event, the same way you would treat a broken tag.

Does GA4 model data for users who decline cookies?

Yes, but only for eligible properties. Google requires the advanced consent mode implementation, at least 1,000 events per day with analytics_storage denied for at least 7 days, and at least 1,000 daily users with analytics_storage granted on at least 7 of the previous 28 days. Modeled data appears under the Blended reporting identity.

Can I use analytics without a cookie banner?

That depends on the law where your visitors are and on what your analytics tool stores, so check your own legal requirements before removing a banner. Google's consent requirements for its tags are written for users in the European Economic Area, and any site with European visitors should assume consent is needed for analytics cookies unless a lawyer says otherwise.

Sources and methodology

Consent mode and GA4 behavior come from the Google Tag Platform and Google Analytics Help pages listed in the sources, read on 9 November 2026. The store in the worked example and all of its numbers are hypothetical; the consent rates, corrected counts and ROAS figures are simple arithmetic done in this article. AdPix facts come from adpix.io and the AdPix docs page on the consent banner.

  1. Google Tag Platform: Consent mode overview
  2. Google Tag Platform: Set up consent mode on websites
  3. Google Analytics Help: [GA4] Behavioral modeling for consent mode
  4. Google Analytics Help: [GA4] Verify and update consent settings in Google Analytics
  5. AdPix Docs: Set up the consent banner

This article was drafted with AI assistance and reviewed by the AdPix team before publishing. Editorial policy

About the authorFounder, AdPix · Builder of AdPix

Founder of AdPix, the web analytics platform built for Iranian teams. Every article on this blog is reviewed by the founder before it is published.

Web analyticsAttributionCampaign measurement

Analyze your data, decide with more confidence

With AdPix, analyze campaign performance and grow in step with your business needs.