Guide
Not one cause. A consent banner, a browser extension, a sampling threshold and a modelled estimate all sit between what actually happened on your site and the number GA4 shows you, and GA4 shows all four the same way: as a number.
Consent banners remove real visitors from the count.
A consent-gated GA4 property only measures visitors who accepted. European acceptance rates commonly leave 40 to 70 percent of real traffic uncounted, and none of it shows up as a gap, because there is no line in the report for "declined".
Ad blockers remove the rest.
uBlock Origin, Ghostery, Brave and a growing share of default browser settings block requests to Google's analytics domains outright. That visitor loads your site, does everything a real visitor does, and GA4 never hears about it.
Sampling replaces a count with an estimate above a threshold.
GA4's standard reports are mostly unsampled, but Explorations, the flexible ad hoc report builder, apply sampling once a query gets large or complex enough. The exact trigger is not fixed, and the report does not always make it obvious when it happened.
Consent Mode fills the gap with a model, not a measurement.
Google's Consent Mode estimates conversions from visitors who declined consent, using a model trained on the visitors who accepted. That is a genuinely useful piece of engineering, and it is also, by definition, not a count of anything that was directly observed.
None of the four things above are hidden by GA4 on purpose. The problem is presentation: an observed pageview and a modelled conversion render in the same font, in the same chart, with the same confidence. A reader has no way to tell, from the screen alone, which number was measured and which was estimated.
Metricward tags every figure with one of four labels, and the label is not decoration, it changes what you are allowed to conclude from the number.
Counted directly from events that arrived. No estimation of any kind.
Exact arithmetic on observed numbers, a rate or a ratio or a sum. Not an estimate, but not a raw count either.
An estimate that fills in something not directly observed, shown with its confidence and the inputs behind it. Never presented as a plain fact.
Written by a system rather than measured, such as an AI summary of a trend. Never rendered as though it were a fact on its own.
A modelled or generated figure always carries its confidence and the inputs that produced it, next to the number, not in a footnote you have to go looking for. If a segment is too small to say anything reliable, the count is shown and the percentage is greyed rather than stated as fact, because a confident-looking "100% conversion" from two visitors has misled somebody at every analytics vendor at least once.
A comparison that wins on every row is not a comparison, it is marketing. These four are real, and worth weighing against everything above.
It costs nothing, at any volume.
GA4 does not meter you by event count. Metricward's free plan stops at 10,000 events a month; past that, it is a paid plan starting at a real price. For a high-traffic site with no privacy pressure and no budget, free wins outright.
It talks to Google Ads directly.
A GA4 conversion feeds Google's bidding algorithms without a second integration. Metricward does not connect to any ad platform, by design, and if your acquisition runs through Google Ads, that is a real cost of not using GA4, not a minor one.
Everyone already knows it.
Every agency, freelancer and marketing hire has GA4 muscle memory. Retraining a team on a new tool has a cost even when the new tool answers a question better.
BigQuery export reaches a scale Metricward does not, yet.
GA4's warehouse export is a mature, unlimited-volume pipe into BigQuery. Metricward exports CSV from every report and from the raw events behind it on every plan, but proper warehouse export is on the roadmap and not shipped.
Deciding whether to move? Read the migration guide for what will not match and the actual install steps, or see the full comparison.
Start the free plan