Put two analytics tools on the same site and they will report different visitor counts. That is normal. Neither one is broken. They are counting slightly different things, with different rules, from requests that do not always reach both.
How big the gap gets depends on your audience and your setup. Here is where it comes from.
Some visits never reach one of the tools
Ad blockers. Blockers work from lists of known analytics addresses. Google Analytics is on every list. Smaller tools are on some. On a site with a technical audience, the share of blocked visitors can be large, and it is rarely the same for both tools.
You can close most of that gap by sending the script through your own domain. See Avoid ad blockers for how we do it.
Consent banners. If your banner holds GA4 back until someone clicks Accept, every visitor who ignores the banner is missing from GA4. If the other tool loads before the choice, or does not wait for it, the two counts drift apart by roughly your decline rate.
Scripts that load late. A tag manager that fires after the page has settled misses people who leave quickly. A script in the page head catches more of them.
Some visits are counted, then removed
Bots. Every tool filters bots, and none of them use the same list. GA4 removes known bots and spiders. Spreelytics also drops traffic from datacenter address ranges when it arrives, which catches scrapers that pretend to be browsers. Ours tends to read lower and steadier as a result.
Modelled data. With Consent Mode turned on, GA4 can estimate the behaviour of people who declined and add it to some reports. Those numbers are filled in by a model rather than counted.
The same visitor is counted differently
What a "visitor" is. Most tools recognise a returning visitor by a cookie. Many cookieless tools work out a fresh id each day instead, so one person across a week is up to seven visitors. Spreelytics can run either way, per audience. See Definitions for how visitors, sessions and people are counted.
Safari. Safari limits cookies written by a script to seven days. Someone who returns after eight days counts as new in most script-based tools, ours included.
Session rules. A session usually ends after 30 minutes without activity. GA4 and Spreelytics both use that default, but GA4 lets you change it, and some tools also start a new session at midnight. Different rules give different session counts from identical traffic.
The report itself is different
Sampling. GA4's standard reports are complete, but its explorations sample once a query covers more than 10 million events on a free property. Thresholds can also hide small rows when Google signals is on. Spreelytics reads every row and has no threshold, so a big range is where the two can differ most. We wrote about how GA4 sampling works.
Time zones. A day in a report set to UTC does not match a day in a report set to New York. Check both tools use the same zone before you compare daily figures.
Date ranges. "Last 30 days" can include today in one tool and stop at yesterday in another.
Revenue disagrees for its own reasons
Your payment provider is the record of what was paid. Analytics revenue can differ from it in a few known ways.
Payments in other currencies are converted, and Spreelytics converts each one at the rate on the day it happened, then freezes it. A payment that arrives without a visitor id still counts toward the total, but it cannot be traced to a channel, so it is listed under Direct. See Revenue for how to pass the id.
How to check a gap
- Pick one week and one page. Compare pageviews first, since they have the fewest rules.
- Make sure both tools use the same time zone and the same date range.
- Look at the split by browser. A gap mostly in Safari points at cookies. A gap mostly in Chrome and Firefox on desktop points at blockers.
- Look at the split by country. A gap mostly in the EU points at your consent setup.
If the pageviews are close and only the visitor counts differ, the tools agree on the traffic and disagree on how to count a person. That is the normal case, and you only need to pick the tool whose definition you trust.