Consent Mode Modeling Checker
Consent Mode only recovers your lost conversions if you clear Google's thresholds — and GA4 and Google Ads use different ones. Enter your traffic, consent rate, and click volume to see which layers qualify, why your two reports disagree, and the two settings that hide modelled data even when you do qualify.
Share granting analytics_storage
Used to convert denied users into denied events
Ads evaluates the threshold per grouping
GA4 behavioural modeling — needs both conditions
Consented users per day
Meets threshold2,200 users/day (needs 1,000 on 7 of the last 28 days)
Users with analytics_storage granted — the model's training data.
Denied events per day
Meets threshold9,000 events/day (needs 1,000 for at least 7 days)
From ~1,800 non-consenting users/day — there must be enough of a gap to be worth modelling.
Google Ads conversion modeling — evaluated per grouping
Ad clicks per country × domain grouping (÷ 3)
Below threshold500 clicks / 7 days (needs 700 per 7 days, per grouping)
Your total is split across groupings before the threshold is applied — this is where most accounts silently fail.
Expect GA4 and Google Ads to disagree.
GA4 qualifies for behavioural modeling but Google Ads does not, so GA4 will report conversions Ads never models. The two layers are independent with different thresholds — this divergence is expected behaviour, not a tagging bug, and chasing it as one wastes a lot of time.
Thresholds are Google's published figures for consent mode modelling and can change; meeting them makes a property eligible rather than guaranteed. Nothing here reads your account — the estimate is computed in your browser from the numbers you enter, and nothing is transmitted.
Why “we have Consent Mode v2” isn't the same as recovering data
Consent Mode v2 is usually installed and then assumed to be working. The banner fires, the signals pass, the implementation is correct — and the conclusion drawn is that the non-consenting traffic is being modelled back in. Frequently it isn't.
Modelling is gated behind volume thresholds, and they are applied in ways that surprise people. Google Ads needs 700 ad clicks per seven days per country and domain grouping, so an international account with healthy total volume can fail in every market at once. GA4 needs two separate conditions met simultaneously — enough denied events to be worth modelling and enough consented users to train on. Miss either and nothing happens.
Then there are the two ways a qualifying property still shows you nothing. If GA4's Reporting Identity is set to Observed, modelled data is stripped out of reports even while it's being generated. And modelled conversions never reach the BigQuery export, which is observed data only — so your warehouse and your GA4 reports will disagree permanently, by design.
Worth being clear about the limits: clearing a threshold makes you eligible, not guaranteed, and models go through a training period before output appears. This tool estimates eligibility from your numbers — it can't see your account, and Google's documentation is the source of truth if the thresholds move.
Frequently asked questions
What are the Consent Mode modeling thresholds?
They differ per platform. Google Ads conversion modeling needs at least 700 ad clicks over a 7-day period, evaluated per country and domain grouping. GA4 behavioural modeling needs at least 1,000 events per day with analytics_storage='denied' for at least 7 days, and at least 1,000 users per day with analytics_storage='granted' on 7 of the last 28 days — both conditions, not either.
Why do GA4 and Google Ads report different conversion counts?
Because they run separate modeling layers with separate thresholds. A property can qualify for one and not the other, and then the platform that models will report conversions the other never recovers. This divergence is expected behaviour rather than a tagging fault — a lot of debugging time gets burned treating it as a bug.
I meet the thresholds but see no modeled data. Why?
Check your GA4 Reporting Identity. When it's set to Observed, modelled data is excluded from reports entirely, even though the modelling is still being generated. Switching to Blended reveals it, and because it's a reporting setting rather than a collection one, the change applies retroactively.
Why is the per-country and domain split such a problem?
Because the Ads threshold is applied after the split, not to your total. An account with 2,000 clicks a week looks comfortably over 700 — until it's spread across five countries, averaging 400 each and qualifying nowhere. Concentrated single-market accounts qualify far more easily than international ones at the same total spend.
Do modeled conversions appear in the BigQuery export?
No. The GA4 BigQuery export contains observed event data only, so your warehouse will consistently show fewer conversions than the GA4 interface once modelling is active. That is by design, not an export failure — build reconciliation expectations around it rather than trying to close the gap.
What can I do if I don't qualify for modeling?
Two things actually move the needle. Raise the consent rate itself — banner design, copy, and layout affect it more than any tagging change. And make the consented data you do collect more durable with server-side tagging and first-party context. Modelling is a backstop for what you lose; it is not a substitute for collecting well.
Does this tool need access to my accounts?
No. It computes eligibility in your browser from the numbers you type. Nothing is transmitted, nothing is stored, and no account access is requested.
Fixing the collection, not just the reporting
When modelling can't reach you, the answer is durable first-party collection: a consent implementation that earns more grants, and server-side tagging that keeps the data you do get.