A sales ad runs for two weeks and barely moves the needle. The owner opens the results, sees a mediocre click rate and no sales to speak of, and has to decide what went wrong. Was it the wrong people seeing it? Too small a budget? Or was the ad itself just not good enough to make anyone stop scrolling? Most owners guess. Few guess right, because the data in front of them usually can't answer the question it looks like it answers.
The instinct is to blame targeting first. It feels like the most fixable problem: pick a different audience, try again. But that instinct assumes the audience you picked is the audience that actually saw the ad. Increasingly, that assumption doesn't hold.
Your targeting settings may not describe who saw the ad.
Jon Loomer, who has run and documented Facebook ad tests for over a decade, spent a year specifically testing how much control advertisers still have over who sees their ads. In one controlled test, he built a custom audience and turned on a Meta feature called Advantage Custom Audience, which lets Meta's algorithm expand delivery beyond the audience an advertiser actually picked. The result, in this single test: only 26% of the budget went to the audience he had actually built and targeted. As Loomer put it, "Since the custom audiences I used for targeting matched how I defined my Audience Segments, we can state definitively that, in this case, Meta spent 74% of my budget reaching people outside of my targeting inputs."
It gets stranger. In a separate test, Loomer ran an ad set with no custom audience, no lookalike audience, and no detailed targeting at all, the kind of setup usually described as going broad and letting the algorithm find people cold. Even then, the platform spent 25% of the budget retargeting people who had already visited his site or were on his email list. Nobody told it to do that. It did it anyway.
The audience report an owner pulls up after a campaign ends may describe almost none of the people who actually saw the ad.
This matters because it means the most common post-launch diagnosis, "we must have targeted the wrong people", is often a guess made with the wrong instrument. The report an owner pulls up after a campaign ends describes the audience that was declared, not necessarily the audience that was reached. If results are flat, the owner has no clean way to tell whether the declared audience was wrong or whether the algorithm quietly routed the budget somewhere else entirely, and the creative simply didn't work on whoever it found. What the owner needs is a way to separate a weak ad from a weak audience before spending more money, which is the judgment you can't make about your own ad, and the one Knowing makes for you.
A weak ad and a misdirected budget look the same from the outside.
Strip away the targeting confusion and there's a second, older problem: a campaign can underperform for reasons that have nothing to do with either the audience or the ad, and everything to do with how the money was spent. Les Binet, whose joint work with Peter Field forms much of the evidence base on how advertising budgets should be split between building a brand over time and driving short-term sales, argued in a 2025 talk that the industry has "become focused on efficiency, targeting, and short-term metrics at the expense of scale, reach, and true brand-building effectiveness." An ad judged purely on short-term numbers can look like a failure when the real issue is that it was never given the reach or the time horizon to do its job.
Binet and Field's separate research on effectiveness in context goes further: the right mix of budget and channels differs by brand and by situation, so a campaign that underperforms might simply have been funded or placed the wrong way for what it was trying to do. None of that is a statement about whether the ad itself was any good. It's a statement about whether the ad was ever put in a position to succeed.
That leaves a business owner staring at the same flat result from three possible directions: wrong people reached, wrong budget or channel decisions, or a creative that genuinely didn't work. The numbers after launch rarely say which.
Even when the ad is the problem, nobody fully agrees on what that means.
Assume, for a moment, that targeting and budget are both fine, and the ad itself is the weak link. Even then, pinning down what went wrong runs into a separate difficulty: advertising researchers don't have a settled definition of what makes creative work in the first place. Arthur Kover, writing in the Journal of Advertising Research, frames it plainly: "Not all advertising, however, is 'creative.' Just what that 'creativity' is and how it makes an impact long has vexed researchers in the fields." That debate hasn't settled in the years since. If the people who study this for a living can't agree on a stable definition, an owner squinting at a bad click rate is not going to reverse-engineer the answer from the outside.
There is at least some evidence pointing at what tends to work better. Kesha Coker and colleagues, in a 2021 study using undergraduate participants, found that video ads built as a story hooked viewers more effectively than ads built as a straight argument, and that being hooked predicted whether someone intended to watch further, share the ad, or talk about it positively. It's worth including precisely because of its limits: this is a lab finding about format and structure, not a verdict on any one campaign. It suggests that whether an ad tells a story or makes a flat argument may matter, but an owner would still need to test their specific ad on their specific audience to know if structure was the reason it struggled, not the targeting or the budget.
The honest fix is to ask the question earlier.
All three of these problems, algorithmic drift in targeting, budget and channel mismatches, and the unresolved question of what makes creative actually work, point the same direction. Once a campaign has run and the money is spent, the data available to an owner is too tangled to cleanly separate a creative failure from a targeting failure from a budget failure. The report shows what happened. It does not show why.
The more reliable moment to ask whether an ad is built to work is before it runs, when the audience hasn't been scrambled by automation yet and the budget hasn't been spent finding out the hard way. At that point, the question isn't "why did this fail," which has too many possible answers, but "is this specific ad, for this specific brand and this specific campaign, built soundly," which has a checkable one. An owner who answers that question before launch doesn't need to untangle Loomer's 74% afterward. They've removed one whole branch of the guessing before the money ever left the account.
Frequently asked
Common questions.
If my Facebook ad underperforms, how do I know if it's a targeting problem or a creative problem?
Often you can't tell for certain after the fact. Meta's automated targeting features can route a significant share of your budget outside the audience you selected, so the post-launch report may not reflect who actually saw the ad. That's why evaluating the ad's structure before launch is more reliable than trying to diagnose the failure afterward.
Does 'going broad' on Facebook ads mean the algorithm shows my ad to anyone?
Not exactly. In Jon Loomer's testing, even with no custom, lookalike, or detailed targeting set, Meta still automatically spent a quarter of the budget retargeting prior site visitors and email contacts, meaning the algorithm applies its own logic even when you supply none.
Why don't advertising researchers agree on what makes an ad 'creative' or effective?
Arthur Kover's research in the Journal of Advertising Research describes creativity in advertising as a contested, unresolved idea among both practitioners and researchers, and that debate hasn't settled since. Without a stable definition of creative quality, it's hard to isolate a creative failure from a media or targeting failure after a campaign runs.