There is a screen inside every ad set you have ever built that you have almost certainly never opened.
It is not hidden, exactly. It sits in the conversion section, underneath the performance goal and the conversion event, behind a gray link that says show more options. Click it and you get one line: attribution setting. Underneath that, a default you did not choose, doing a job you were never told about.
Most owners who find it assume it is a reporting preference. A way of deciding which numbers show up in the dashboard. Something for the accountant.
It is not. That setting is one of the few places in Ads Manager where you tell Meta's delivery system what kind of customer to go and find. Leave it alone and you have still made a choice. You have just made it silently.
The Setting Almost Nobody Opens
Meta lets you choose an attribution model at the ad set level. For most accounts, the model is Standard, and the setting underneath it reads 7-day click and 1-day view.
In plain terms: if somebody clicks your ad and converts at any point in the next seven days, Meta counts it. If somebody sees your ad without clicking and converts within one day, Meta counts that too.
Two things follow from that, and only one of them is widely understood.
The first is the reporting half. Those are the conversions that appear in your columns. This is the half that gets written about, usually in the context of Meta claiming credit for sales your bank account cannot find, which is a real problem and one we have already taken apart in detail.
The second half is the one that costs money quietly.
It Is an Instruction, Not a Report
Meta does not just count conversions inside that window. It optimizes delivery to produce them.
That is the whole thing. The attribution setting is read by the delivery system as a definition of success. Meta looks at the window you left in place, works out what kind of person tends to convert inside it, and then goes looking for more of those people. The window is not a lens you look through afterwards. It is a brief you hand over beforehand.
Once you see it that way, the default stops being neutral. A 7-day click and 1-day view window tells Meta that a person who saw an ad yesterday and bought today is a success worth repeating. For a lot of businesses that is exactly right. For others it is an instruction to go and spend money on people who were already going to buy.
What Each Window Actually Counts
There are three standard settings, and Meta's own documentation is precise about each one.
Click-through counts conversions that happen within one day or seven days of somebody clicking a link in your ad. Seven days is the ceiling. If you have read an older guide promising a 28-day click window, that guide predates the 2021 privacy changes and is describing a control that no longer exists.
View-through counts conversions within one day of somebody seeing your ad without clicking it.
Engage-through counts conversions within one day of a non-link click. That means any interaction with the ad that is not a click through to your site, and for video it means the viewer watched five seconds, or ninety-seven percent of the video if the video runs shorter than five seconds.
One caveat worth carrying, in Meta's words: some accounts may still use prior versions of click-through and engage-through attribution while the current version rolls out. If your panel does not look like the description above, that is why.
Right for a Purchase, Wrong for a Simple Lead
Here is where this becomes worth twenty minutes of your afternoon.
The default is built for a considered purchase. Somebody sees a pair of boots, thinks about it, mentions it to someone, comes back on Thursday and buys. That journey does not run in a straight line and it does not finish in an hour, so a seven day click window is a fair description of how the decision actually happened. Leave it alone.
Now take a different conversion. You are a local service business. Your ad offers a free guide, or a quote, or a price list, in exchange for an email address. That is not a considered purchase. That is a thirty second decision. Nobody sees your ad on Monday, reflects on it for six days, and then decides on Sunday that yes, they would like the PDF.
If you leave the default in place on that ad set, you are telling Meta that a week-long path to an email address is a normal and desirable outcome. Meta believes you. It widens its net accordingly.
For a straightforward lead, the setting should be one day click only. No view window, no seven day tail. The conversion is immediate, so the instruction should say so.
The third case is the one that does real damage, and almost nobody checks it. If you optimize toward a custom event, something like time on page, or a scroll depth, or an internal click, ask yourself whether a returning visitor could trigger that event over and over. Because if they can, and you leave a view window switched on, Meta will do precisely what you asked. It will find the people most likely to fire that event repeatedly, which is your existing audience, the people already visiting your site. Your cost per result will look magnificent. Your new customer count will not move.
That is not Meta misbehaving. Meta is not guessing at what you meant. It is reading the instruction literally and executing it well.
Look Before You Touch
None of this means you should go and start changing settings this afternoon. There is a free step first, and it does not alter delivery at all.
In Ads Manager, open the columns menu and choose Compare Attribution Settings. Select a few windows at once. You will get the same ad sets reported several ways side by side, and you will see, in your own account rather than in an article, how many of your conversions arrive on the first day versus how many drift in across the week.
That breakdown answers the question for you. If almost everything lands inside one day, your seven day window is not buying you accuracy, it is buying you a wider and vaguer instruction. If a real share of conversions arrives on day four and day five, the longer window is describing something true and you should keep it.
There is one rule to respect while you are in there, and it comes straight from Meta: results cannot be compared in the campaign overview table across ad sets using different attribution models, because each model counts differently, and comparing across them will lead to inaccurate conclusions. Use the compare feature. Do not eyeball two ad sets running different models and conclude one is beating the other.
This is also the cheapest possible version of the broader discipline of not trusting a single platform's self-reported numbers, which Meta has structural reasons to inflate.
The New Option: Incremental Attribution
Standard is no longer the only model. Meta now offers two others.
Incremental attribution is the interesting one. In Meta's own words, it is an attribution model that optimizes ad delivery for incremental conversions, using machine learning models that predict whether a conversion is caused by an ad.
Read that again, because the important word is optimizes. This is not a reporting view. Switch it on and Meta changes who it shows your ads to. It tries to identify the people who would not have bought without seeing the ad, and shifts delivery toward them, and away from the people its models think were going to convert regardless.
For any owner who has ever looked at a retargeting report and thought that half of these people were already coming back anyway, that is the itch this scratches. It is the logic of a holdout test, which we have written about as a discipline, built into the delivery system and running continuously.
There is a fence around it. To see the option at all you need a website conversion location, a sales campaign, and a purchase or catalog optimization, with a performance goal of maximize number of conversions or maximize value of conversions.
Which means if you run lead campaigns, and most local and service businesses do, this is not available to you yet. That is not a workaround situation. The setting will not appear. It is reasonable to expect lead objectives to follow eventually, but today it is a purchase-led feature.
You Can Look Without Switching
This is the part worth knowing even if the fence above rules you out, and it is the single most useful thing on this page.
Meta's documentation is explicit that you do not have to adopt the model to see its numbers: even if incremental attribution is not your attribution model, you can still view modeled results based on it. In Ads Manager, open the columns menu, choose Compare attribution models, select Incremental, and apply. Nothing about your delivery changes. You are simply shown the same campaigns counted a second, stricter way.
That is the cheapest honest read you will ever get on whether Meta is claiming credit it has not earned. Meta's own worked example: a campaign with 100 total conversions, 70 of which it considers incremental, shows 70 in the incremental column against 100 under standard attribution. The gap is the portion of your reported performance that Meta itself does not believe its advertising caused.
Two limits worth carrying. The data does not exist for any date range before 1 April 2025, so there is no history to look back on. And Meta recommends comparing within the same attribution setting rather than across models, because each one counts differently. Read the gap as a signal about direction and scale, not as a precise subtraction.
The third model is Custom attribution, which lets you feed attribution data from your own analytics tool back to Meta so delivery optimizes against your logic rather than Meta's. It is genuinely powerful and it is genuinely not for a business without an analytics team. Know it exists. Move on.
Your Numbers Will Drop. That Is the Point.
If you are eligible and you do switch an ad set to incremental attribution, brace for the dashboard.
One agency published the same ad set counted three ways: 473 purchases on a 7-day click window, 417 on 1-day view, and 330 on incremental attribution. Same ads, same spend, same week. Those are three different counting methods rather than three measurements of one thing, so the arithmetic between them is not exact. The scale of the spread is the point. The lowest number is not a failure. It is the same business results measured against a stricter definition of what your advertising actually caused.
This is the part that catches people out, and it catches them out in an expensive direction. An owner switches the model, sees the purchase count drop by a third, panics, and switches back within a week. Nothing was broken. The ruler changed.
Meta's documentation is careful here and it is worth matching its restraint. It says incremental attribution can help you focus campaigns on outcomes more likely to be directly driven by an ad, and potentially drive more incremental conversions than your existing campaigns. You will find much bigger numbers quoted in videos and in sales conversations, and they do not agree with each other, which is reason enough to treat the mechanism as the useful part and the percentages as marketing.
When to Leave All of This Alone
There is a case for touching none of it, and it applies to more readers than the sections above.
Narrowing an attribution window narrows the data flowing back into the ad set. On an account with healthy conversion volume, that is a fair trade: you lose some quantity of signal and gain a lot of quality. On an account producing a handful of conversions a week, it is not a trade at all. You are taking a system that was already short of data and giving it less.
The same caution applies with more force to incremental attribution, because genuinely incremental conversions are a subset of an already small number. Feed a low-volume ad set the strictest possible definition of success and it may never gather enough of them to settle, which is a slower and more confusing failure than simply leaving the default in place.
So the honest sequence is this. If your ad sets are not yet clearing enough conversions to get out of the learning phase, attribution settings are not your problem and adjusting them will not become the answer. Volume comes first. Come back to this screen when you have something to measure.
What to Actually Do About It
Open one ad set. Find the attribution setting under show more options and write down what it says, because you have probably never looked.
Then go to the columns menu, run Compare Attribution Settings, and look at where your conversions actually land across the windows. That costs nothing and changes nothing.
Then match the instruction to the decision. A considered purchase keeps the default. A quick lead should almost certainly be one day click. A custom engagement event should be one day click, or you will fund a very efficient campaign to reach people who already know you.
That is most of the value on this page, and none of it requires new creative, a bigger budget, or a call with anyone.
If you would rather somebody who looks at these panels every day went through your account and matched every ad set's attribution setting to the decision it is actually asking people to make, that is part of what we do when we run Facebook and Instagram campaigns for small businesses. Either way, go and open the setting. It has been giving Meta instructions in your name the entire time.
