Metrics and analysis
Attribution windows: why the same month has two numbers
7-day click, 1-day view, click date or conversion date. What each choice changes in the report, and which one to use when deciding where money goes.
Also available in: Português · Español
Two reports for the same month, from the same account, with different conversion numbers. Nobody made a mistake. The attribution window is different — and this is one of the few places in paid media where two contradictory truths coexist without either being a lie.
The two questions a window answers
How long afterwards does it still count? Someone clicked the ad on Monday and bought on Friday. Is that a sale from the ad? With a 7-day window, yes. With a 1-day window, no.
What counts as contact? Clicking is obvious. What about seeing the ad without clicking and buying the next day? That is view-through attribution, and it is far looser.
That is why a window is written as two numbers: 7-day click, 1-day view is the most common default. For the plain definition and what changes between 1, 7 and 28 days, see attribution window.
The table that explains most arguments
Same account, same month, different windows:
| window | conversions | CPA |
|---|---|---|
| 1-day click | 180 | $233 |
| 7-day click | 247 | $170 |
| 7-day click + 1-day view | 310 | $135 |
| 7-day click + 7-day view | 412 | $102 |
No row is false. The "real" CPA ranges from $102 to $233 depending on a setting — which is why an agency and a client can argue for an hour about who is right when both are looking at different settings.
The practical rule: pick one, write down which one in the report, and never compare periods using different windows.
The detail that confuses most: click date
The platform records the conversion on the click date, not the date the conversion happened.
Someone clicked on 28 September and bought on 2 October. With a 7-day window, that sale appears on 28 September — a day you already read, already reported and already closed.
Two effects follow that alarm anyone who does not know:
Yesterday's number changes today. That is normal. The window is still open.
A closed month keeps rising. If you exported on the 1st, the last days of the month will keep collecting conversions for another week. A monthly report closed on the 1st always understates the end of the month.
The fix: close the monthly report with at least 7 days of slack, or state that the final days will still move.
Why Meta and Analytics never agree
They are different models solving the same problem in incompatible ways.
| Meta | Google Analytics | |
|---|---|---|
| Attributes to | whoever saw or clicked in the window | last non-direct click |
| Records on | click date | conversion date |
| Sees | only what passed through Meta | only what arrived with tracking |
| Cross-device | yes, via login | partial |
If someone sees an ad on Instagram, searches on Google and buys, Meta claims it (seen within the window) and Analytics gives it to Google (last click). Both count the same sale, for different owners.
Adding the two together is the most common and most expensive mistake: you end up with more conversions than there are sales.
Which window to use when deciding money
To compare creatives against each other: any window, as long as it is the same for all of them. The comparison is relative; the absolute value does not matter.
To decide budget: use the most conservative window that still gives you volume. If you decide using 7-day view, you are crediting the ad for sales that might have happened anyway.
To talk to finance: none. Use CRM sales divided by total spend. It is the only number that survives an audit, because it depends on no model at all.
That last one is the calculation described in true cost per sale, and it is what belongs at the top of a client report.
The sanity check nobody runs
Take a closed month and compare:
conversions the platform claims ÷ real sales in the CRM
If the platform claims 310 and the CRM recorded 180 sales in the period, the ratio is 1.7. That is not fraud — it is the sum of view-through attribution, a long window, and conversions that would have happened without the ad.
Knowing that factor is useful: it becomes your mental discount. When the platform says 100 next month, you already know the business will see about 60.
Run this calculation once a quarter. It shifts when the window changes, when the campaign mix changes, and when somebody touches tracking without telling you.
To sum up
The window is not a technical detail, it is an editorial choice about what you consider the ad's merit. Pick one, declare it, and never compare periods on different windows.
And keep beside it the number that depends on no window at all: CRM sales divided by spend. That is the one the business owner feels in the bank account.
If the gap between the ads manager and the CRM is your problem right now, see when the CRM and the ads manager disagree.
Frequently asked questions
Which attribution window should I use?
7-day click and 1-day view is the sensible default for most accounts. What matters is picking one and always comparing against it, because changing the window changes the entire history.
Why did yesterday's number change today?
Because the platform attributes the conversion to the click date, not the conversion date. A sale today from someone who clicked five days ago lands on the click day, changing a number you already read.
Does view-through attribution really count?
It counts loosely. The person saw the ad, did not click, and converted later. For remarketing it usually inflates results considerably, so look at click and view separately.
Why does Meta show more conversions than Google Analytics?
Different models. Meta claims the conversion from anyone who saw or clicked within the window; Analytics attributes by last click. There is no perfect reconciliation between the two.