Debrief · 11 October 2026 · 7 minutes
Why GA4 and Google Ads never agree, and which one to believe
- Google Ads and GA4 both report real numbers. They attribute the same month differently.
- Five settings decide what each tool counts. Together they are a lens, and in most businesses it was never chosen.
- Choose the lens for the decision in front of you, write it down and mark every gap unreconciled until it has a reason.
Demo data. CobberBank is a fictional bank. Every figure in this piece is a September 2026 snapshot of NR's demo room. None of it comes from a client.
Both numbers are real. In the September 2026 snapshot of NR's demo room, Google Ads counts 7,114 paid search conversions and GA4 counts 3,284, for one bank, one month and one set of campaigns.
This is not a tracking problem. It is a lens problem. Neither tool is broken. Each counts different things, at different moments, under different rules, so the lens you read through matters as much as the number itself. In most businesses that lens is a decision nobody has made.
Here is the whole gap for paid search.
| September 2026 snapshot, paid search only | Google Ads, search campaigns | GA4, Paid Search channel |
|---|---|---|
| Visits | 170,298 clicks | 116,897 sessions |
| Conversions | 7,114 | 3,284 |
Five reasons explain a gap like this.
Five reasons and the five checks a GA4 audit starts with
A GA4 audit starts with these five. Call it a Google Analytics audit if you prefer.
1. A click is not a session
Google Ads counts a click the moment someone clicks the ad. GA4 counts a session when the page loads, the tag fires and consent allows it. Some clicks never become sessions. The visitor left before the page loaded. The click went to a phone call instead of the site.
Not every Google Ads click is a paid search click either. In the September 2026 snapshot, CobberBank's Google Ads account recorded 439,364 clicks. Only 170,298 came from its two search campaigns. The rest came from Display, Demand Gen and Performance Max, which GA4 files under other channels. Compare search with search.
The quieter loss is the click identifier. A session only counts as paid if that identifier survives the trip to the site. A redirect, a missing tag or a consent setting can strip it. The visit still happens. GA4 files it under another channel. In the September 2026 snapshot, across every paid channel, GA4 credited one in twenty of the sessions that arrived on a paid click to a paid channel. Three in five sat in Unassigned. Just over one in three sat in Organic Social.
It can run the other way too. GA4 files a returning direct session under that visitor's last paid or organic click. It does the same with any conversion in that session. That is how paid sessions can exceed clicks.
The fix. Check that auto-tagging is on, that the Google Ads link in GA4 is live and that no redirect drops the click identifier.
2. The attribution windows differ
Google Ads dates a conversion to the day of the click. GA4 dates it to the day it happened. A click on 29 September that converts on 3 October is a September conversion in Google Ads and an October conversion in GA4.
The windows differ as well. By default Google Ads looks back 30 days for a click. GA4 looks back 90 days for most key events. Google Ads also sees only its own ads, so it credits itself. GA4 sees every channel and shares the credit.
The fix. Record the window and the attribution model in each tool, then compare like with like. Google Ads has a "by conversion time" column that dates conversions the way GA4 does.
3. Consent and modelling fill the gap differently
When a visitor declines cookies, both tools lose sight of them. Both then estimate what they missed. Google Ads estimates conversions. GA4 estimates visits and events.
So part of each number is observed and part is estimated. The estimated parts have no reason to agree. The lower your consent rate, the wider the gap.
The fix. Know your consent rate. Check that Consent Mode is set up the same way for both tools. Note which reports include modelled data.
4. Time zone and currency
Each tool keeps its own time zone and its own currency. Put the Google Ads account on one clock and the GA4 property on another and a day's numbers shift by hours, a month's at both ends. If the currencies differ, every value is converted at a rate you did not choose.
CobberBank's accounts share one time zone and one currency, so this reason adds nothing to its gap. Check it first anyway. It is the only one of the five you can rule out in two minutes.
The fix. Match the settings where you can. Where you cannot, state the offset beside the number.
5. "Conversion" means two different lists
Google Ads counts the conversion actions marked as primary. GA4 counts the events marked as key events. These are two lists, kept in two places, often by two people.
CobberBank's Google Ads account counts one thing as a conversion: an application. It values each one at A$550. That value is an assumption about what an application is worth. It is not revenue. So the A$3,912,700 of conversion value its search campaigns show in the September 2026 snapshot is 7,114 applications multiplied by an estimate.
The fix. Put the two lists side by side. For each line write what it counts, how often it counts and what it is worth. That page is a definitions file. Every number after it means the same thing to every reader.
Conversion settings are the first of twelve checks on a paid account. The rest are in Google Ads audit: the checks that find money, and the one that finds the truth.
Which one to believe
Believe each tool for the job it was built for.
| The decision | The number to use | Why |
|---|---|---|
| Bids and budgets inside Google Ads | Google Ads conversions | It is the number the bidding sees. |
| Comparing one channel with another | GA4, once its channels are reconciled | One ruler across every channel. |
| What marketing delivered, for finance and the board | The CRM or the finance system | It holds the outcome the business is paid for. |
| Whether something changed this week | Either tool, against its own history | Each tool is consistent with itself. |
Size changes the answer as well. A smaller business can run on GA4 and keep softer conversion points, such as a form started, as a secondary measure of whether the month is tracking to target. Above the mid-market line, a larger advertiser tends to move to an ad server such as Google's Campaign Manager 360, which brings its advertising into one shop and counts it under one set of rules. Each way of reading the numbers has pros and cons. The costly choice is the one left unmade.
Trouble starts when one number is asked to do every job.
What a reconciled view looks like
A reconciled view does not pick a winner. It puts the numbers side by side, shows the gap and states the reason for it.
| September 2026 snapshot, paid search only | |
|---|---|
| Google Ads clicks, search campaigns | 170,298 |
| GA4 sessions, Paid Search channel | 116,897 |
| Clicks with no matching paid search session | 53,401 |
Some of the 53,401 never became a session. Some became a session that GA4 filed under another channel. A reconciled view says how many of each, with the reason beside it. In this snapshot the 53,401 are still marked unreconciled. A budget set on either tool alone is a budget set in the dark.
Three rules make a view reconciled:
- Every number carries its definition.
- Every gap has a stated reason. Until it has one, it is marked unreconciled.
- The total never changes. Sessions move between channels. They are never invented.
What to ask before you hire a Google Analytics consultant
Whether the work is a GA4 audit run in house or a brief for a Google Analytics consultant, six questions come first:
- Is auto-tagging on and is the Google Ads link live?
- How much traffic sits in Unassigned and Direct? What is in it?
- Which key events exist, how is each one counted and which are imported into Google Ads?
- Which attribution model and which lookback windows does each tool use?
- Do the time zones and currencies match?
- What share of visitors consent? Where does modelling fill in?
Ask for the answers on one page, with the size of each gap beside its cause.
That page is where a Number Check starts. In three weeks, with read-only access, NR reconciles the last twelve months across your platforms, your analytics and your CRM.