Every ad platform counts its own results, and platform reported conversions are the figure that comes out. Meta counts what Meta saw. Google counts what Google saw. Neither is lying, and adding the two together still gives you a number that exists nowhere in your business.
Platform reported conversions are a claim about cause, produced by the party whose budget depends on the answer. That does not make them useless. It makes them a specific kind of evidence with specific limits.
This is for you if your ad reports and your sales figures have never quite agreed.
What platform reported conversions actually count
They count events the platform recorded, that it can connect to an interaction it served, inside a window it chose.
Every part of that sentence does work. The platform can only count what reached it, so blocked events and refused consent vanish. It can only connect what carries an identifier. And the window decides how far back a click can reach to claim credit.
Change the window and the same week produces a different number. Nothing about your business changed. The measurement did.
That is the honest description of these figures. They are not sales. They are the platform’s best account of its own contribution, made with partial information.
Why platform reported conversions come out high
Four things push platform reported conversions above what a finance system would recognise, and they compound.
Both platforms can claim the same sale. Somebody clicks a Meta ad on Tuesday, searches your name on Thursday, clicks a Google ad and buys. Each platform saw a real interaction and each credits itself.
View-through counting. Somebody who never clicked but was shown an impression can be counted if they convert inside the window. Sometimes that is genuine influence. Often it is a coincidence with a budget attached.
Long windows. A conversion credited to a click from three weeks ago is credited on the assumption the click caused it, which grows less likely the further back you reach.
And returns never arrive. The platform records the purchase and never learns it came back, so your reported revenue quietly overstates the money you kept.
How far apart the numbers get
The honest answer is that it varies more than most articles admit, and anybody quoting you a single multiplier is guessing.
Seer Interactive tested Meta’s incremental attribution setting across $1.05 million of ad spend on six accounts in April 2025. Meta reported 87% of conversions as incremental. On the same accounts, over the same period, GA4 put it at 67%. Twenty percentage points apart, from two systems watching the same events.
In my own accounts the gap between platform reported conversions and what the client’s CRM records is usually around 10%. A typical week might show the platform reporting 50 leads against 45 in the CRM.
That is smaller than this genre usually claims, and I think the difference is worth saying out loud. Most writing on this subject implies platform numbers are inflated by multiples as a rule. In lead generation, with tracking set up properly, they mostly are not. A small number of my clients do run at two or three times, and those accounts have a specific fault rather than a general condition.
So the useful question is not “how inflated is it” but “how inflated is mine, and why”.
The gap is a fingerprint, not a fault
A stable gap is genuinely useful information.
If your platform reports 10% more than your CRM every month, you can work with that. Apply the correction, judge trends inside the platform, and report the CRM figure to the business. Nothing is broken.
What matters is stability. A gap that moves around means something is changing that you have not identified, and that is when it becomes worth an afternoon of investigation.
So measure the gap before you try to close it. Most people attempt to fix a discrepancy they have never actually sized.
The one habit that keeps platform reported conversions honest
Compare like with like, on the same window, or do not compare at all.
This sounds obvious and it is broken constantly. Somebody pulls Meta on a 7-day click window, Google on a 30-day window, and the CRM on calendar months, then puts all three in one table. The table looks authoritative. It compares three different questions.
Fixing it costs an hour, once. Decide the window you report on, set every platform to it where you can, and note where you cannot. Some settings are not adjustable, and knowing that is better than assuming they match.
Then check whether view-through is included in what you send upward. Many people have never looked, and it is usually a column you can add to the report rather than a project. On a retargeting-heavy account it can be a large share of the claimed total, and it is the component least likely to survive a controlled test.
The point of all this is not perfect measurement, which is not available. It is knowing which parts of your number are solid and which are the platform’s opinion, so that when somebody senior asks, you have an answer rather than a shrug.
Where platform reported conversions belong, and where they do not
This is where most of the damage happens, and it is a communication problem rather than a technical one.
Use platform reported conversions for daily and weekly steering inside a single platform. Which campaign is improving, which creative is working, where the budget should go tomorrow. They are good at this, because the bias is roughly constant across the campaigns you are comparing.
Use a system the business already trusts for anything that reaches finance. Revenue, growth, whether marketing pays for itself.
Never add two platforms’ claimed conversions together and call the total sales. That single habit is responsible for more lost credibility than any other reporting mistake I see.
What to change this week
Three steps.
Pull one month of platform reported conversions and one month from your CRM, matched to the same window, and write down the gap as a percentage. That number is more useful than any benchmark.
Then check whether view-through conversions are included in what you report upward. Many people do not know either way, and it is usually a column you can add.
Then agree, in writing, which system the business judges marketing on. Not which is right, which is the one you all quote.
What conversion numbers actually measure sets out the mechanism, the four ways the number goes wrong covers the failure modes, and sending closed-won data back is how you stop the platform optimising towards the wrong people in the first place. Fixing this properly usually means changing what the site sends, which is web development work as much as advertising work. If you want the gap in your own account sized and explained, book a teardown.