It is Tuesday morning. You open Google Ads and it tells you last month produced 61 conversions. You open GA4 and it tells you the same month produced 40. Same business, same website, same thirty-one days. Somebody is wrong, and you are about to lose an afternoon working out who.
Nobody is wrong. The two products are not disagreeing with each other, because they were never in the same conversation. They were built to answer different questions, they count on different days, they wait different lengths of time, and one of them is allowed to estimate. Every one of those differences is documented by Google itself. None of them is a defect.
The instinct is to treat the gap as a problem to be solved. It is not. It is a property of the tools, the way the difference between your fuel gauge and your fuel receipts is a property of trucks. One tells you what is in the tank. The other tells you what you spent. Both are true, neither is the other, and averaging them would be a strange thing to do.
What you actually need is not two numbers that match. It is knowing which number answers which question.
They Were Built for Different Jobs
Google Ads exists to decide what to bid on. That is its entire purpose. Every conversion figure it reports is in service of one question: which clicks are worth paying more for tomorrow?
GA4 exists to describe your website. Everything that happens on it, from every source, whether you paid for the visitor or not.
Same business, different job, different answer. Once that lands, most of the panic goes away, and the remaining differences stop being mysteries and start being settings you can go and look at.
Google Ads Files the Sale on the Day of the Click. GA4 Files It on the Day of the Sale.
This is the difference that catches people first, and it is the one nobody guesses.
Google's documentation is blunt about it. The primary conversion columns "are calculated based on the time of the click, not the time of the conversion." If someone clicks your ad on Monday and books on Thursday, Google Ads records that conversion on Monday, next to the click it paid for. GA4 records it on Thursday, when it actually happened.
For a plumber, a law firm, a roofer, anyone with a few days of thinking time between the click and the phone call, this alone will make any short date range look broken. Pull last week in both platforms and you are not comparing two answers to one question. You are comparing two different sets of events that happen to share a date range.
It also explains the spike you see when you launch a promotion. Google Ads shows a wall of conversions on launch day, because that is when people clicked. GA4 spreads the same sales across the following week, because that is when people actually bought.
There is a fix for this that almost nobody uses. Google Ads has a parallel set of "by conv. time" columns that report on the date the conversion happened instead of the date of the click. Add them to your view and this entire category of confusion disappears, because now both platforms are filing on the same day.
Google Ads Only Sees Its Own Clicks. GA4 Sees Everyone Who Touched You.
A customer finds you in organic search in early March. They think about it. A week later they click your ad. They sit on it, get your follow-up email, and finally call.
Google Ads sees the click it sold you and credits the campaign. That is the only touch in the journey it has any visibility into.
GA4 sees the whole path and divides the credit, because in its default configuration for Analytics reports it attributes across paid and organic channels, not just Google's. Organic found the customer. Paid closed them. Email nudged them over. Three contributors, one sale, credit shared.
Google Ads is not inflating and GA4 is not deflating. One has a narrower field of view than the other, on purpose, because it is only ever being asked about the traffic you paid for.
Worth knowing if you go looking at attribution settings: the menu is smaller than it used to be. First click, linear, time decay and position-based models were all retired in November 2023. What remains is data-driven, paid and organic last click, and Google paid channels last click.
Two Settings Nobody Opens Decide a Chunk of the Gap
Both platforms have a counting setting. Most owners have never opened either, and the combination they happen to be running can produce a large gap all by itself.
In Google Ads, every conversion action is set to count either Every or One. Every counts every conversion that follows an ad interaction, which is what you want for sales, because a customer who buys three times is worth three times as much. One counts a single conversion per ad click, which is what you want for leads, because the same person filling in your form twice is not two customers. One is already the default for calls from ads, calls to a number on your website, and imported Analytics goals.
In GA4, every key event is set to count either Once per event or Once per session. Once per event is the recommended setting and counts the event every time it fires. Once per session counts it once inside a session no matter how many times it happens.
Here is the part worth checking today. Once per session is the default for every key event that was created from an old Universal Analytics goal. If your property was migrated rather than built fresh, there is a good chance you are running the legacy setting without ever having chosen it.
Run Google Ads on Every and GA4 on Once per session and Google Ads will report a higher number permanently. That is not a measurement truth about your business. That is two dropdowns disagreeing.
You Are Probably Comparing the Wrong Column
If you only act on one thing in this article, make it this one, because it accounts for more inflated Google Ads numbers than anything else on this list and it takes about ten seconds to check.
Google Ads has a Conversions column and an All conversions column. They are not the same thing and they were never meant to be compared to the same GA4 report.
Google's wording is explicit: view-through conversions "are not included in the 'Conversions' column, only in the 'View-through conversions' and 'All conversions' columns." A view-through conversion is someone who was shown your ad, never clicked it, and later converted anyway.
All conversions also carries conversion actions you deliberately set to secondary and chose to keep out of your main reporting, store visits, and calls placed after someone viewed your ad on a tablet or a computer.
So if you have been comparing GA4 against All conversions, you have been comparing your website's key events against a column that intentionally contains people who never clicked, actions you already decided not to optimize toward, and visits that happened at your physical door rather than on your site. Of course it is bigger. It is supposed to be bigger.
One Platform Waits Longer for the Sale Than the Other
Every conversion has a deadline attached to it, and the two platforms set theirs differently.
Google Ads gives you a click-through conversion window of 1 to 30, 60 or 90 days on Search and Display campaigns, and it defaults to 30 days. Its view-through window defaults to a single day, and the engaged-view window to three.
GA4 runs a lookback window of 30 days for acquisition key events and 90 days for everything else, with 30, 60 and 90 available.
Read those side by side and the trap is obvious. On the standard defaults, GA4 will keep crediting a conversion back to a source for ninety days while Google Ads has already closed the book at thirty. Google's own documentation notes that conversions "can be reported up to 90 days after the click." For a business where people decide in an afternoon this barely matters. For a roofer, a dentist or anyone selling something people save up for, it matters a great deal.
One Platform Estimates. The Other Counts What It Can See.
This is the difference that is growing, and the one most likely to be misrepresented to you by somebody selling a fix.
When a visitor declines cookies, Google does not simply lose them. Consent mode modeling uses what it can observe to estimate what it cannot, and Google states plainly that those "modeled conversions will appear in the 'Conversions' column," integrated with "the same granularity as observed conversions." Google also publishes the reason the estimate is not a simple scale-up: "consented users are typically 2-5x more likely to convert than unconsented users."
GA4 does its own version, modeling the behavior of users who decline analytics cookies from the behavior of similar users who accept.
So one of your two numbers contains estimated conversions and the other contains a different set of estimates, built from different inputs, for a different purpose. They were never going to land on the same integer. The same mechanic is why Meta's reported sales run ahead of your bank; it is not a Google quirk, it is how modern ad measurement works.
You will find a lot of confident percentages online for exactly how much GA4 undercounts. Be careful with them. Nearly every figure in circulation traces back to a company that sells server-side tracking, and the ones that do cite a source tend to cite each other. The direction is real and well documented. The decimal places are marketing.
When the Gap Is a Bug and Not a Design
Everything above is working as intended. Sometimes your tracking genuinely is broken, and it is worth knowing what that looks like so you do not spend a month calmly explaining away a real defect.
The common ones: you are tracking a purchase in one platform and a thank-you page view in the other, which are not the same event and will never reconcile. Duplicate tags are firing, so one platform counts twice. Auto-tagging is off or the conversion linker is broken, so neither platform can match a click to the visit that followed it. Or a third-party checkout is carrying the wrong conversion ID, which Google specifically calls out for platforms like Shopify.
You will see thresholds quoted for this, usually along the lines of a quarter being normal and a half being suspicious. Google publishes no such number, and the honest test is better than any threshold anyway. Go down the list above and try to name which cause accounts for your gap. If you can point at the column, the counting setting, the window and the date basis and say "that is where it comes from," it is design. If you get to the bottom of the list with nothing to point at, go looking for a defect.
The other tell is shape rather than size. Differences by design are stable and boring. They sit there month after month at roughly the same proportion. A real break usually announces itself as a sudden change, or a zero where a number used to be.
There Is a Third Number, and It Is the Only One That Pays You
Google Ads says 61. GA4 says 40. Your calendar says you booked nine jobs.
Nine is the only number in that sentence that bought anything. Both dashboards are counting proxies for revenue, and neither of them knows what happened after the phone rang. A form fill from someone outside your service area, a wrong number, a tire-kicker who was never going to sign, and your best job of the quarter all land in the conversion column as one apiece.
None of this is unique to Google. No ad platform can tell you exactly which ad made the sale, because causation is not a thing a platform can observe. What is worth your attention is the narrower, fixable version of the problem: feeding real outcomes back into the account, so the bidding is at least chasing the kind of lead that turns into work. Your books are the source of truth on money. The dashboards are instruments for steering.
Which Number for Which Decision
Here is the whole thing, reduced to the only part you have to remember.
Making decisions inside Google Ads? Use the Conversions column in Google Ads. Not All conversions, and not GA4. That column is the exact number Smart Bidding is optimizing against. If you judge a campaign on any other figure, you are grading it on evidence it was never shown.
Deciding whether paid is worth it next to organic, email and referrals? Use GA4. It is the only one of the two that can see the other channels, and channel-mix questions are the job it was built for.
Asking whether the money is actually working? Use your books. Closed jobs, signed contracts, deposits taken.
One number, one job. The moment you stop asking all three sources the same question, they stop contradicting each other and start being useful.
That is also the difference between an account that gets managed and one that gets worried about. We spend our days inside these dashboards for small businesses running Google Ads, and the accounts that do well are rarely the ones with the tidiest reporting. They are the ones where somebody decided, once, which number they were steering by.
The goal was never a matching number. It is being able to point at the gap and name exactly where it comes from. That is a much shorter afternoon than the one you were about to lose, and it is worth considerably more.
