Not every weak campaign fails the same way. But after reviewing roughly a hundred of them, most of the failures fall into one of four patterns. It's almost boring how often we see the same shapes.
If you've been running ads for a while, you'll probably recognize at least two of these in your own account. We did, in ours.
1. Campaign sprawl
The single most common pattern: too many campaigns doing almost the same job.
A team launches a "main" campaign, sees mixed results, launches a slightly different one to "test", forgets to pause the first, then layers in two more variations a month later. Six months in, the account has 18 campaigns, and three of them are quietly responsible for 80% of the spend. The other fifteen are noise, but they're consuming budget the optimizer thinks is "exploration."
The fix isn't dramatic. Audit, consolidate, archive. Most accounts can lose 30 to 50% of their campaigns and improve performance the same week.
2. Broad messaging with no defined intent
The ads read like a brand brochure. The targeting is "people who might like this." The copy could swap into a competitor's account and nobody would notice.
This usually traces back to the briefing stage. Whoever briefed the campaign didn't have a sharp answer to "who specifically, and what specifically are they trying to solve right now." So the campaign hedges. Hedged campaigns get hedged results.
The cheap test: read your three best-performing ads and your three worst, with the brand name removed. If you can't tell which ones are yours, neither can the algorithm.
3. Testing without a testing logic
Every account claims to be testing. Most aren't.
Real testing has a hypothesis ("we think X will outperform Y because Z"), a single variable changed, a sample size that can give a real answer, and a decision rule for what happens when the test concludes. Most "testing" we see has none of these. It's three ad variants thrown into the same group, the platform picks one, and the team writes a Slack message that says "headline B is winning."
That's not a test. That's the platform serving whichever variant got a few early clicks. The "winner" is mostly noise, and a month later you can't replicate why it won, because no one wrote down what made it different.
4. Reporting that doesn't support decisions
Beautiful dashboards. Six tabs. Nine charts. And when you ask "based on this report, should we increase spend on Campaign X?", nobody can answer cleanly.
Reports have to point somewhere. If a metric doesn't trigger an action when it moves, it doesn't belong on the dashboard. Half the metrics we see on most ad reports are decorative. They exist because the platform exposes them, not because anyone uses them to decide anything.
A useful exercise: for every chart on your dashboard, write down what action you'd take if the metric moved 30% in either direction. The charts where you can't answer that question are the ones to delete.
What this means
Bad performance is, almost always, a systems issue before it becomes a media-buying issue. The ad isn't underperforming. The structure around the ad is.
That's also good news. Systems issues are fixable. They don't require a creative breakthrough or a hidden growth hack. They require an honest audit, a willingness to consolidate, and a refusal to keep dashboards that don't drive decisions.
Most accounts get back 20 to 40% performance just by removing what shouldn't be there. We've seen it too many times to call it a coincidence.
KaiNet · Campaign patterns