Estimator for Google Analytics 4
The traffic GA4 doesn’t count.
Visitors who decline or ignore the cookie banner don’t show up in your reports. Upload an export, enter your consent rate and get an estimate of the totals, with the range of uncertainty written next to it.
With a 60% consent rate, GA4 counts about 69 of 100 visits, likely between 61 and 77: the boundary is blurred because it depends on the assumptions. The others, about 31, are without consent. The simple calculation would say 60.
- simple calculation
- GA4 sees 60 visits out of 100, 40 are missing.
- estimate
- GA4 sees about 69 visits out of 100 (likely between 61 and 77), about 31 are missing.
Example with 100 visits: visitors who accept come back 3 times on average, those who don’t, 2.
How it works
Three steps, nothing to install
No account and nothing to install: everything happens in this browser page, and the file never leaves your computer.
- ≈ 3 min the first time
Upload the export
A CSV from a GA4 free-form exploration: sessions, key events and transactions by day and channel. How to create it
- ≈ 1 min
Answer a few questions
How Consent Mode is set up and what your banner’s consent rate is. “I don’t know” is a valid answer: it widens the range.
- a few seconds
Read the estimate and refine it
A central estimate with its range. Then you can change the assumptions, add real orders from your order system and see what changes.
What you get
A number, how far to trust it, and where it comes from
These are the estimator’s real results, calculated on the sample data: the same cards and charts you’ll see with your own export.
Sample data (fake), calculated with the estimator
GA4 shows you about 65% of sessions.
Likely between 54% and 75% (8 times out of 10), if the assumptions are right.
Why 65% and not 60%? 60% is the share of visitors who accept. Visitors who accept come back more often (assumption: those who don’t accept make 0.8 visits for each of theirs), so they weigh more in sessions.
And about 82% of transactions (between 68% and 91%): the visits you don’t see convert less, so proportionally fewer orders are missing.
Sample data (fake), calculated with the estimator
By channel · Sessions
Sample data (fake), calculated with the estimator
Which assumption weighs most on the uncertainty of sessions, and how much the total would change if it were wrong.
Where the uncertainty comes from: Sessions
- consent rate 30%
- how often they come back 68%
- little data 1.8%
What if an assumption were wrong?
Sample data (fake), calculated with the estimator
The CSV, first rows and assumptions
section;item;metric;observed;estimated_total;likely_min;likely_max;almost_certain_min;almost_certain_max summary;"Total";"Sessions";38211;58882.0;51048.5;71099.4;48145.7;80573.0 summary;"Total";"User-days";37228;57373.0;49733.8;69274.2;46907.9;78505.2 summary;"Total";"Key events";2203;2677.5;2386.0;3420.1;2317.0;4179.1 … 228 more rows by channel and by day … assumption;value;range_min;range_max;status "Consent rate used";0.599;0.502;0.692; "How often they come back";0.800;0.444;1.440;"Default" "How well they convert";0.400;0.150;1.066;"Default" "How much they spend per order";1.000;0.745;1.342;"Default"
- CSV
- PDF with header, date and assumptions
- Link with only the assumptions
CSV and PDF are created in this tab. They always list the assumptions used and how many are defaults.
What it solves
Dividing by the consent rate isn’t enough.
Two separate effects shift the count. The example is the same across the page: 100 visits, a 60% consent rate, visitors who accept make 3 visits and those who don’t make 2.
Visitors who accept come back more often
6 visitors out of 10 accept, but GA4 counts 18 sessions out of 26.
That’s why the share of sessions you see is higher than the consent rate.
Visitors who don’t accept buy less
The missing visits hold about 2 orders, not 4.
Out of 100 visits GA4 counts 69, with 9 orders. The 31 without consent convert about 0.4 times as often.
Illustrative example
That’s why dividing isn’t enough
In the example the simple calculation overestimates sessions by 15% and orders by about 40%. Uncounted accounts for both effects and tells you how uncertain the estimate is.
What it is not
What it doesn’t do, up front
It isn’t a measurement
It’s an estimate that depends on stated assumptions: change them and it changes. That’s why every number comes with its range.
It covers consent only
Traffic GA4 misses because of ad blockers or tracking errors isn’t included.
It isn’t needed if GA4 is already estimating
With Consent Mode Advanced and Google’s modeling active, the estimator notices and stops, so nothing is counted twice.
It also assumes opt-in consent, where tracking starts only after the visitor accepts, as in the EU and the UK. Under opt-out rules, common in the United States, the missing share is usually much smaller: more in the Guide.
How the method works · How it was tested
Who made it
Uncounted is a project by Davide Gelatti, Digital Analyst (opens in a new tab). It was built with Claude, Anthropic’s AI model, in phases, with every decision approved by the author. How it was made
How much traffic is your GA4 missing? Find out in five minutes.
Your file stays in your browser. No export at hand? Start with the sample data.