I want you to do something before you read another word of this. Pull up your ad account, find the attribution setting, and look at what window it's set to. Most of you are going to see "7-day click, 1-day view." Now go pull your actual bank deposits or Shopify orders for the same period and compare them to what the platform says it drove.
They won't match. They're not supposed to match. And the gap between those two numbers is the whole reason your CAC "looks fine" while your cash position tells a different story.
The window isn't measuring what you think it's measuring
Attribution windows are not truth. They're accounting rules a platform picked for its own reasons, and those reasons have almost nothing to do with your actual payback period. The windows themselves are arbitrary — there's no scientific reason why 7 days is the right timeframe for one platform or 30 days for another. These are business decisions made by platforms, not natural laws of consumer behavior. Your customers don't know or care that Meta gave itself a 7-day credit window and Google gave itself 30.
Here's the mechanism, plainly: Meta's attribution settings determine which conversions appear in your Ads Manager reports — select a 7-day click window, and your reports include every conversion that happens within seven days of someone clicking your ad. On top of that, Meta's default is still 7-day click plus 1-day view, meaning a conversion gets attributed to your ad if the user clicked it within 7 days before converting, or simply saw it within 1 day before converting without clicking at all.
That view-through piece is where most of the fake efficiency hides. View-through credits a purchase to an ad someone scrolled past without clicking, and a lot of that credit actually belongs to organic, email, or direct traffic instead. One industry estimate puts it plainly: view-through attribution adds 20-40% to reported conversions but represents partial inflation. Another source citing Meta's own research states 67% of advertisers using default settings over-attribute conversions by 15-30%, leading to budget misallocation and inflated ROAS calculations.
I'm not going to pretend every one of these stats comes from a peer-reviewed source — some of this is agency-published research, and you should treat the exact percentages as directional, not gospel. But the direction is consistent across every source I looked at: the default window flatters you, and it flatters you predictably.
Why "it collapsed" is not a tracking bug
This isn't hypothetical anymore, either. Meta itself has been narrowing the runway. Meta removed 7-day view and 28-day view from Ads Manager in January 2026, and while 28-day click is still accessible via the API for some accounts, the longest window available to most advertisers is now 7-day click plus 1-day view. Then in March, Meta redefined what even counts as a "click" and split social interactions into a separate engage-through category. The result, according to multiple agency writeups: these changes caused reported conversions to drop 15-40% overnight across millions of accounts — not because performance declined, but because the measurement methodology fundamentally shifted.
Read that again. Nothing changed about how many people bought your product. What changed is how generous the platform's own scoreboard was willing to be. If your CAC suddenly looked worse this year, before you fire your media team, check whether you're actually looking at a reporting change dressed up as a performance change.
The real number lives somewhere else
The honest way to find your real payback period is to stop asking the platform to grade its own homework. That's what incrementality testing is for — conversion lift testing measures how many sales your ads actually caused, not just correlated with. You hold out a portion of your audience, don't show them ads, and compare what happens.
The gaps this uncovers are not small. One incrementality research firm's dataset found that across 640 incrementality experiments, Meta drove roughly a 19% average lift — while retargeting's incremental ROAS runs 40-70% below platform-reported numbers. Read that as: the platform is telling you retargeting is your best channel, and a controlled test is telling you it's mostly taking credit for people who were already going to buy.
A separate write-up walked through this exact scenario on a live account: independently measured, first-party, click-only attribution put the true return at 0.93x — the gap between platform-reported and real performance, made concrete on a single account. A platform telling you 3x and reality delivering under 1x is not a rounding error. That's the difference between a scalable channel and a business slowly bleeding out while the dashboard says everything's fine.
What I actually tell clients to do
I'm not telling you to abandon platform reporting — you still need it day to day to know if a campaign is trending up or down. But you cannot use it as your only input on whether to scale. Here's the process I run, and it works whether you're a DTC brand or a B2B SaaS account with a longer sales cycle:
- Pull the platform number and your backend number for the same 90-day window. Meta or Google conversions on one side, Shopify orders or CRM-logged deals on the other. The delta between them is your inflation rate, full stop.
- Run the math on view-through specifically. If view-through conversions make up more than a quarter of what's reported, that's a flag, not a footnote — a widely cited rule of thumb across attribution research is that if view-through conversions are more than 25% of your total reported conversions, your ROAS is probably inflated.
- Test a narrower window against your backend data before you scale, not after. Duplicate the campaign, set one to the default window and one to click-only, run both for two weeks, and see which one's number actually lands closest to what your bank account says happened.
- Run an actual holdout when the budget justifies it. You don't need a six-figure lift study to get directional signal. Even a geo holdout or a platform-native conversion lift study on a meaningful chunk of spend will tell you more truth than another quarter of dashboard-watching.
- Separate your optimization window from your evaluation window. Let the algorithm learn on the wider window if that's what it needs to find buyers. But grade the campaign, and make the scale-or-kill call, on the number that matches your backend.
Why this matters more than most people admit
Every brand I've watched scale into trouble did it the same way: dashboard says efficient, so they pour more budget in, and the platform happily finds more of the same "conversions" — a chunk of which were never going to be incremental in the first place. The algorithm isn't lying to you maliciously. It's optimizing toward the definition of success you gave it, and if that definition is generous, it will keep finding generously-defined wins.
This is exactly why attribution and incrementality are not competing ideas, they're different tools for different jobs. Attribution is the day-to-day dashboard — useful, fast, directionally helpful. It is not the number you should be using to decide whether next quarter's budget goes up or down. That decision needs the slower, less flattering number: what actually happened to revenue when the ad wasn't there.
Your CAC probably does look great. I'd just want to know great compared to what — the platform's version of events, or the one your business is actually living inside.