iubenda With Causality Engine: Check What GA4 Can See

The short answer: before asking which channel sold, ask how many sales GA4 saw at all, because a consent banner decides that first. iubenda, the generator of privacy policies and cookie banners with Google Consent Mode built in (entry), installs on Basic Consent Mode by default: Google services stay blocked completely until the visitor consents (auto-blocking guide). A visitor who refuses sends GA4 nothing, and a purchase GA4 never received cannot be attributed by any method. Causality Engine, which reads the store's GA4 export, works inside that same limit, so measure the gap first and then read the file; the first finding is free on the start page.

Disclosure: I founded Causality Engine and own it, and this page recommends it. Every claim about it links to Causality Engine's own pages.

Three questions, in this order:

1. What share of orders does GA4 hold? For example, 1,540 GA4 purchases against 2,200 paid store orders in the same weeks is 1,540 / 2,200 = 70%. That one ratio limits every report built on GA4, attribution included.
2. Which Consent Mode is the site on? Open the site's settings in iubenda, the tab Block scripts prior to consent, and look at "Do not block Google's services that adhere to Consent Mode". Unticked means Basic; ticked means Advanced (auto-blocking guide).
3. Is GA4 modelling anything? In GA4, the reporting identity and the data-quality icon say whether estimated data is in your reports (behavioral modeling help).

iubenda and the read: what each controls

The number iubenda puts forward, and what it rests on

iubenda sends no traffic and claims no sales. The result it points to is recovered measurement. Its Consent Mode page says Consent Mode recovers conversion data on around 65% of ad clicks from users who did not consent, and that Google's models fill in behaviour for people who rejected marketing consent. The same page describes the two modes: Basic turns Google tags off until consent; Advanced preloads them and sends cookieless pings when consent is declined.

Two facts from the vendor's own guides set the scale of that claim:

• Recovery needs Advanced, and Advanced is not the default. New installations start on Basic, with Google services blocked completely until consent (auto-blocking guide). Advanced is a box you tick, and the guide warns that IP addresses may still reach Google when consent is refused, so it needs careful consideration for privacy compliance.
• Measuring the consent rate is a top-tier feature. Detailed analytics, covering consent rates, button clicks and notice closures, appears in the Ultimate column of the pricing page, at €79.99 a month billed yearly (checked on 11 November 2026).

What GA4 records under each setting

Google's consent mode help describes the denied state: tags store no cookies and send cookieless pings with coarse information instead, and Google Analytics fills some gaps with modelling. That only applies when the tags load. On Basic they do not, so a refuser leaves no trace.

The modelling has conditions of its own. Google's behavioral modeling help requires consent mode on all pages, tags that load before the dialog and in all cases, at least 1,000 events a day with analytics_storage denied for 7 days, and at least 1,000 daily consenting users on 7 of the previous 28 days. Below that, events from refusers are not reported. Even when modelling is on, Google lists data export among the features that do not use modelled data, and most exploration templates show consenting users only.

So for a store, the banner settles three things before any attribution model runs: how many purchases exist in GA4, whether any are estimated, and whether the paths behind them are whole. Channel labels are the same either way; what shrinks is the number of journeys carrying them.

A worked example: two consent setups, one month

For example, with numbers invented for illustration: a store books 2,200 paid orders in March. Under Basic, GA4 records 1,540 purchases. Suppose, purely as an assumption for the sum, that visitors arriving from paid social refuse more often than those arriving by email:

In the store, Paid Social is 800 / 2,200 = 36.4% of orders. In GA4 it is 480 / 1,540 = 31.2%. Email goes the other way, 18.2% in the store and 328 / 1,540 = 21.3% in GA4. No attribution rule produced that shift; the banner did, before the first touchpoint was counted.

A causal read cannot put the 660 missing orders back. What it does with the 1,540, per the how-it-works page, is split Direct back to the channels behind those buyers, show each channel's caused share next to a last-click rebuild, grade data health from 0 to 100, and suggest a next step tested in one region first. The one-off read page is blunt about the floor: no causal method recovers data that was never collected. If the read and an ad platform disagree about a channel with a low consent rate, test that channel in one region before trusting either.

The free check: the ratio, weekly, by hand

No tool and no upgrade needed. Each Monday, write down two counts for the previous week: paid orders from the store's admin, and purchases from GA4. Divide. A ratio that holds steady is a known blind spot; one that drops after a banner change, a theme update or a new tag manager is a measurement fault, not a sales trend.

Then run the refusal test iubenda's own guide warns you to run, because its auto-blocking "remains experimental, and its accuracy may not be flawless" (auto-blocking guide): a private window, refuse everything, and watch the browser's network tab for requests to Google. On Basic there should be none until you accept.

What the read leaves to iubenda, and the reverse

Consent is iubenda's job and the counsel's, not the read's. Whether Advanced is lawful for your audience is a legal decision. The read in turn needs a GA4 purchase key event and 12 months or more of history (how it works), and a channel under 1% of credit on both methods is called too small to act on.

iubenda's page sells Consent Mode as a way to keep conversion tracking and get more out of an ads budget, and the figure it quotes, 65% of non-consented ad clicks recovered, is a figure about Google Ads measurement on the setting the installer does not switch on. The banner is metered by pageviews whichever mode you pick (pricing page), so the recovery claim costs iubenda nothing to make and is yours to verify.

Then read what GA4 holds

Once the weekly ratio is known and steady, export GA4's Attribution paths report and drop it on the start page; the first finding runs in your browser and is free. Causality Engine, which I founded, sells the full read of that file for €99 once, excluding VAT, refundable within 30 days (pricing, checked on 11 November 2026). Read its channel shares with the ratio beside them.

See also

• iubenda: plans, pageview tiers and the free plan's limit
• How to use iubenda: blocking trackers until the visitor consents
• iubenda on the marketplace: iubenda's affiliate program terms
• Causality Engine: the read, its inputs and refusals
• How to use Causality Engine: export, read and test one region
• Causality Engine on the marketplace: program terms, given as the owner's
• Consent-friendly EU analytics: GA4 behind a banner, with a second counter
• Complianz vs iubenda: two consent tools from one group

FAQ

Does iubenda pass consent data to Causality Engine?

No. Nothing connects them. iubenda tells Google tags what the visitor chose; Causality Engine reads the GA4 export that results, as its integrations page describes, with whatever gaps the banner left.

Will switching to Advanced Consent Mode fill the gap in the export?

Not by itself. Advanced sends cookieless pings, and GA4 models from them only above Google's thresholds; Google lists data export among the features that do not use modelled data (behavioral modeling help). Decide the mode on legal grounds, not for a fuller export.

Can the read correct for visitors who refused?

No. It works under the same consent limits as GA4 (how it works). Keep the weekly ratio of GA4 purchases to store orders beside its results so the blind spot stays visible.