Why your analytics and your ad platform disagree, and which to believe

Google Analytics says one number. Your ad platform says another. The gap is often 20 to 40 percent, and no tool is broken. Two systems are measuring different things, and both call the result a conversion.

Almost every company that spends on ads has lived this moment. You open the ad platform and it reports, say, 487 conversions. You open Google Analytics and it shows 312. Same month, same campaigns, two confident numbers that cannot both be right. Someone asks which one goes in the report, the room debates it for twenty minutes, and the meeting ends with a shrug and a decision made on a number nobody quite trusts.

The instinct is to assume one tool is broken. Usually neither is. They disagree because they were built to answer different questions and to count in different ways, and once you understand the handful of reasons why, the disagreement stops being a mystery and becomes something you can manage. You will never make the numbers match exactly. You can absolutely build one number you are willing to defend.

01They are answering different questions

Start with the deepest reason, because it explains most of the gap. An ad platform is built to prove that its ads worked. Google Analytics is built to describe what happened on your site across every channel. Those are not the same question, so they were never going to produce the same answer.¹

The ad platform asks how many conversions its ads drove, and it counts generously, because generosity is in its interest. Analytics asks how many conversions happened on your site and what contributed to each, and it splits the credit across everything involved. When the same sale gets full credit in the ad platform and a fraction of the credit in analytics, you get two different totals from one event, with neither lying.

02The three mechanics behind the gap

Underneath that, three concrete differences do most of the damage. None of them are exotic, and seeing them by name is usually enough to end the Monday argument for good.

Different attribution windows. Each platform decides how long after an ad it will still claim a sale. Meta, by default, counts a conversion that happens within seven days of a click or one day of a view.² Google Ads windows are configurable and can stretch much longer. Analytics applies its own lookback. The same purchase can fall inside one platform's window and outside another's, so it gets counted by one and ignored by the other.²

Views versus clicks. This one is large. Meta will credit a conversion to someone who merely saw an ad and bought within a day, with no click at all, because its logic is user-based and influence-based rather than session-based.² Google Analytics does not track ad impressions or view-through conversions at all, so a sale Meta proudly claims as a view-through is, to analytics, simply a visit from some other source.¹ One platform counts the influence of being seen; the other only counts the click.

Different clocks and different counting. Google Ads records a conversion on the date of the ad click, while Google Analytics records it on the date the conversion actually happened, which can be days later.³ Ad platforms can also count every conversion from one person, while analytics often counts one per session.² Stack these together and a 20 to 40 percent disagreement between an ad platform and analytics is now considered a normal range, not a sign that anything is wrong.⁴

Here is how ordinary that gets. A customer clicks a Meta ad on a Monday, gets busy, and buys the following Tuesday, nine days later. Meta's default seven-day click window has already closed, so Meta records nothing for that sale.² Google Analytics, working from a longer lookback, still credits the original source and counts it. Same purchase, opposite treatment, purely because of a window setting that neither you nor the customer ever saw. Multiply that across a month of real buying behavior and the totals drift apart on their own, with no error anywhere.

03Privacy widened the gap

The disagreement got worse after 2021, and it is worth knowing why, because it will not reverse. When Apple let iPhone users block cross-app tracking, the large majority opted out, and platforms like Meta lost much of the view-through signal they used to connect an ad impression to a later sale.¹ To fill the holes, the platforms now model conversions they can no longer see directly, which means a growing share of those confident platform numbers are estimates rather than counts.

Analytics has its own version of the problem. When visitors decline tracking, modern consent rules can leave Google Analytics undercounting paid conversions by a substantial margin in regions with high rejection rates, with some analyses putting it in the 20 to 40 percent range.³ So one side increasingly models up, the other increasingly misses down, and the gap between them stretches further. Neither is the truth. Both are partial.

One side increasingly models conversions up, the other increasingly misses them down. The gap is not a mistake to fix. It is two partial views you have to reconcile.

04So which one do you believe?

The honest answer is neither, at least not on its own. An ad platform only sees its own corner and is motivated to claim credit. Analytics is more neutral but blind to impressions, limited in its models, and dented by consent loss. Picking the bigger number because it flatters the campaign, or the smaller one because it feels safe, is just choosing which bias to adopt.

The useful move is to stop trying to crown a winner and start reconciling. Compare the platforms over the same window, expect a known and stable gap between them, and investigate only when the gap suddenly changes, because a moving gap is the real signal that something broke. The goal is not one tool that is right. It is a method that gives you a consistent, defensible read on whether a campaign is actually working.

A worked example makes that concrete. Suppose your ad platform reports 480 conversions and your analytics reports 320 in the same month. The platform is running about 50 percent ahead. If it ran roughly 50 percent ahead last month and the month before, that gap is just the structural difference between the two tools, and you can plan around it with confidence. If it suddenly jumps to 90 percent, something has changed: a pixel firing twice, a broken tag, an edited window. That break is what deserves your attention, and you would never have spotted it if you were chasing a perfect match instead of watching a stable ratio.

05Building the one number you can defend

The durable fix is to create a layer that sits above the platforms and owes loyalty to none of them. That is what a single source of truth is for, and the pieces are well understood.

The last row is the one that ends the argument permanently. When you export both your ad data and your real orders into a single warehouse and join them on campaign and date, you stop asking which platform to trust and start measuring against the only number that actually pays the bills, which is the revenue in your own system.³ The platforms become inputs to optimize against rather than competing versions of reality.

One habit to drop along the way: adding the platform numbers together. If Meta claims 200 conversions and Google Ads claims 180, you did not earn 380. The two are often claiming the same buyers, since a person can see one ad and click another, and summing across channels is exactly how companies talk themselves into believing they sold far more than the bank account shows. Credit belongs to the customer journey, not to every platform that happened to touch it. Your warehouse, joined to real orders, is what tells those apart.

You do not have to build all of it at once. Aligning windows and tightening UTMs alone closes a real part of the gap in an afternoon. But the reason the disagreement never fully goes away is structural, and the only lasting answer is a measurement layer you own. That is precisely the foundation everything else we do sits on, because every decision about budget, channels, and growth rests on whether you can trust the number underneath it.