GA4 data discrepancies: why your numbers differ and how to reconcile them

GA4 data discrepancies are normal: two analytics tools never count the same traffic identically, and GA4 will not match your ad platforms or your order database either. The differences come from about eight known causes, including how sessions are defined, which bots are filtered, how declined consent is handled and which tags get blocked. What matters is whether the gap is stable. A steady gap is a definition difference you can live with; a gap that moves from one week to the next means something in your tracking broke.

Data discrepancy is the difference between two systems’ counts of what should be the same thing, such as page views in two analytics tools or purchases in an analytics tool and in the store’s backend. It is usually expressed as a percentage of the number you treat as the reference.

Why do GA4 numbers differ from other analytics tools?

GA4 and any second tool disagree because they make different choices at every step between a visitor’s browser and your report. Some choices remove traffic, some split or merge it, and some only shift it between days or channels. Knowing which direction each cause pushes tells you where to look.

Bot filtering is a good example of how invisible these choices are. Google says known bots are identified “using a combination of Google research and the International Spiders and Bots List” and that “you cannot disable known bot traffic exclusion or see how much known bot traffic was excluded.” A second tool with a different list, or none, will count more page views, and neither report tells you how many.

Consent is the other large one. Google’s behavioral modeling “uses machine learning to model the behavior of users who decline analytics cookies,” but only for properties with at least 1,000 events a day with analytics storage denied for 7 days, and at least 1,000 daily users who granted consent on 7 of the previous 28 days. A small site never qualifies, so declined visits simply disappear from its GA4 users. A large site may see modeled users that no other tool has.

Which metrics should you compare first?

Start with the metric both tools define the same way and move toward the ones they define differently. Page views are the simplest thing any tag counts. Users are the hardest, because each tool identifies people with its own cookies, IDs and modeling.

1. Page views on a few key pages. Home page, a category page, the checkout page. If these disagree a lot, the tags are not installed the same way, and nothing further down the list is worth comparing yet.
2. Specific events. Add to cart, form submit, sign-up. Both tools should fire them on the same trigger.
3. Conversions against your backend. Purchases or leads in each tool divided by the count in your order database or CRM. This is the only comparison with a real ground truth.
4. Sessions by channel. Expect gaps here from session rules, referral settings and attribution.
5. Users. GA4 shows active users in reports, which Google defines as users who “engaged with your site or app,” while total users counts anyone who “triggered any event.” Another tool may count every visitor. Treat this comparison as the least informative.

Write the metric definitions down before you look at numbers. Comparing a tool’s “visitors” to GA4’s active users and calling the gap a discrepancy is the most common mistake in these tests.

How do you run a 30-day side-by-side comparison?

Run both tags on every page for at least 30 days, compare settled data weekly, and judge the gaps by their stability rather than their size. Two weeks set the baseline; the remaining weeks show whether it holds.

Before day one:

• Install the second tag on every template, including checkout and thank-you pages. A tag manager container makes this one change instead of many.
• Set both tools to the same time zone and currency.
• Give conversions identical names and triggers in both tools.
• Tag every campaign link with the same UTM scheme. Mixed casing such as Instagram and instagram becomes two sources in most tools.

Each week: pull the same date range, ending at least three days ago, from both tools and from your backend. Record each metric and the gap as a percentage. A spreadsheet with one row per week and one column per metric is enough.

A worked example

A hypothetical online store runs GA4 and a second analytics tool side by side for four weeks. The backend records 300, 310, 290 and 300 orders, 1,200 in total.

Read it in two passes. The page view gap stays between −5.8% and −6.3% every week, so it is a stable definition or filtering difference and needs no fix. The purchase share is the interesting part. GA4 captures 87% of backend orders in weeks 1, 2 and 4 and 70% in week 3 (203 ÷ 290), while the second tool stays near 85% throughout. One tool moved on its own, which points at a GA4-specific change that week, such as a checkout redesign that dropped the purchase tag. In week 4 the share returns to 87%, so the fix worked.

Notice what the totals hide. Over the full month GA4 caught 82.9% of orders and the second tool 85.5%, which could be read as “the second tool is more accurate.” It is not: in three normal weeks GA4 caught slightly more. The weekly view shows a three-day bug, while the monthly view would have produced a wrong conclusion about the tools.

Why don’t Google Ads clicks match GA4 sessions?

Google Ads counts clicks on ads and GA4 counts sessions on your site, so they were never meant to match. Google’s troubleshooting page lists the usual reasons, and most of them are not errors.

• Invalid clicks. “Google Ads filters out invalid clicks, while Google Analytics initially reports all sessions before potential filtering.”
• Repeat clicks. “A user might click your ad multiple times in a short period,” which is several clicks and one session.
• Returning users. People who come back during the campaign can add sessions without new clicks.
• Bookmarks. A bookmarked landing page URL with its click ID records sessions with no click.
• Slow pages. A visitor who leaves before the tag runs is a click with no session.
• Technical issues. “GCLID stripping, incorrect tag implementation, or session ID changes on the website.”

Conversions differ by date as well as by count. Google Ads “attributes conversions to the date of the ad click,” while GA4 reports the conversion event on the day it occurs. A purchase on Monday from a click the previous Friday lands in two different weeks.

If clicks and sessions diverge sharply for one campaign only, check its landing page first: a redirect that drops the query string removes both the click ID and the UTM tags.

Which number should you trust?

Trust the system that owns the event. Your backend owns orders and revenue, the ad platform owns spend, impressions and clicks, and the analytics tool owns behavior: which pages, paths and channels led to the orders. Decide this before the comparison, so nobody argues later about which tool is “right.”

Then use the analytics tool for proportions, not totals. If an analytics tool captures a stable 87% of backend orders, its split of those orders by channel is still useful, and you can report revenue from the backend and channel shares from analytics. Our website tracking guide shows how to plan events and purchase fields so that share stays high.

When is a discrepancy a real problem?

A discrepancy is a real problem when it changes suddenly, when it affects one page, browser or channel far more than others, or when it changes a decision. A stable 6% gap in page views changes nothing. A conversion rate that is 2% in one tool and 3% in the other changes which campaign gets the budget.

Work through the usual suspects in this order:

1. Tag coverage. Is the tag on every template, especially checkout and pages built by a different team or plugin?
2. Duplicate firing. A tag installed both in code and in a tag manager doubles page views in one tool only.
3. Redirects and parameters. Does any redirect drop UTM tags or click IDs?
4. Sampling and thresholds. Is the GA4 number from a sampled exploration? Our guide to GA4 data sampling explains the 10 million event quota and how to spot it.
5. Late data. Did you compare days that had not finished processing? Google says GA4 “ignores events that arrive more than 2 calendar days, plus today” late.

If the comparison is part of choosing a replacement for GA4, add your own criteria to the gap check: which reports your team opened, which ones were missing, and how much work the setup took. The Google Analytics alternative guide gives a decision table for that choice, and the Matomo vs Google Analytics comparison covers the self-hosted option.

Where does AdPix fit in a side-by-side test?

AdPix runs alongside Google Analytics without interfering, so it can be the second tool in exactly this kind of test. AdPix is a web analytics platform that replaces Google Analytics 4: realtime, acquisition, engagement, conversion and e-commerce reports, plus attribution, lift testing and marketing mix modelling, with no data sampling and a free plan.

For the comparison itself, three facts matter. Install takes adding the AdPix code or importing a GTM container, so the second tag can go live on every page in one change. AdPix reports are not sampled, which removes one of the eight causes from the table above. And the Free plan includes 100,000 events a month with 30 days of raw data, which suits a small site’s 30-day test; a busier site should check its monthly event count first, since Starter raises the limit to 1 million events.

Two features from the Starter plan up are relevant to discrepancies specifically: invalid traffic detection, for teams that suspect bots or click fraud behind a gap, and built-in consent management, so the consent banner and the measurement come from the same product. The AdPix features overview lists what each plan includes. AdPix data is hosted in Iran; readers elsewhere should check their own legal requirements on data location before adding any second tool.

Before you remove anything, keep a copy of your history. The guide to exporting GA4 data covers report downloads, the Data API and BigQuery, and lists what cannot be recovered later.

Start this week: write down the five metrics you will compare and their source of truth, put the same UTM scheme on next week’s campaigns, and record your first settled week as the baseline.

UTM link builder

Tag campaign links so every click is credited to the right channel. The free plan covers 100,000 events a month, no credit card. Setup takes under five minutes.

Start freeTry it free