For years the skill in Meta advertising was building the audience. Layered interests, behaviours, lookalikes at three different percentages. That skill has quietly stopped mattering, and Meta ads audience targeting is now the least useful lever in most lead generation accounts.
What replaced it is your creative. The ad selects the audience: whoever responds tells the system who to look for next. Spend your time there instead.
This is for you if you still build audiences the way you did three years ago.
What Meta ads audience targeting used to do
The old logic was reasonable. Tell the platform who your customer is, and it shows your ad to those people.
So you would stack interests, layer in behaviours, build a lookalike from your customer list, and exclude the obvious. The person who assembled the cleverest audience got the cheapest leads, and that was a genuine craft.
It worked because the system needed the help. With limited signal, your guess about who to reach was better than its guess.
That condition no longer holds. The system now has far more signal than you do, and your guess is mostly a constraint on it.
Why broad beats hand-built Meta ads audience targeting
A broad audience gives the algorithm the widest pool to learn within.
It shows your ad, watches who responds, and adjusts. Every conversion narrows its picture of who is worth reaching, and it does that continuously, at a scale of detail you could not specify by hand.
Narrowing removes people it might have found. Worse, small audiences exhaust: the same people see the ad repeatedly, frequency climbs, response falls, and your cost per lead rises. That looks like creative fatigue and is really audience starvation.
I have watched accounts improve simply by deleting the interest stack and letting the system work. Not because broad is magic, but because the constraint was costing more than it saved.
Meta ads audience targeting that still earns its place
Three things survive, and they are worth doing properly.
Exclusions. Do not pay to reach people who already bought, and exclude recent enquirers so your figures are not flattered by people already in your pipeline. This is the part of targeting that has not been automated away.
Geography. The system cannot know you only serve three counties. Getting this wrong is the most expensive simple mistake in the whole account.
Age, but only where it is genuinely required, such as a licensing or legal restriction. Tightening age bands for performance reasons usually costs reach without improving quality.
Everything else, I would test removing before I would test refining.
Creative is doing the Meta ads audience targeting now
If the ad selects the audience, then the ad is where your effort belongs.
A different creative angle reaches a different group of people, without you changing a single setting. Show the same service through the eyes of a worried customer rather than a bargain hunter, and the system finds worried customers.
That is a more powerful lever than any interest stack, and it is the one people spend the least time on. Most accounts I open have twelve audiences and three ads. It should be closer to the reverse.
It also means creative volume matters. Not more of the same, but genuinely different reasons to care, each of which opens a different pocket of the audience.
Why narrow targeting costs more than it used to
Wasted reach was always inefficient. It is now expensive in a way it was not three years ago.
Meta reported average price per ad up 12% year over year for the quarter ending 30 June 2026, with impressions up 14%. So the platform is selling more inventory at higher prices.
That changes the maths on a narrow audience. When you restrict the pool, you burn frequency on a small group, and every one of those repeat impressions now costs more than it did. The penalty for over-constraining has gone up while the benefit has gone down.
There is a second effect worth naming. Rising prices make people tighten their targeting, because narrowing feels like the responsible response to costs going up. It is usually the opposite of what the account needs, and it is how a bad quarter becomes a bad year.
What this changes about testing
Old testing compared audiences with the same ad. New testing compares ads with the same audience.
That is not a small change to your process. It changes what you brief, what you produce, and what you measure. If your monthly rhythm is built around audience experiments, you are running the test that stopped being informative.
It also changes what a losing test means. A creative that fails has not just lost a comparison, it has told you that a whole group of people does not respond to that framing. That is useful, and audience tests rarely produced anything that transferable.
What I would actually do to an account today
Delete the interest stacks and run broad, keeping exclusions and geography.
Give it two weeks and enough budget to leave the learning phase, because judging before then means reading noise.
Put the time you were spending on audiences into producing three genuinely different creative angles a month, not three variations of one.
Then judge on cost per qualified lead rather than cost per lead, because broad can bring cheap volume that converts poorly when the creative is vague about who it is for.
One thing I would not do: rebuild everything at once. Change the targeting or change the creative approach, not both in the same fortnight, or you will not know which produced the result. Sequencing two changes costs you a couple of weeks. Confusing them costs you the answer entirely, and you will end up guessing for the rest of the quarter.
What to change this week
Four steps.
Open your best-performing campaign and note how many audiences and how many distinct creative angles it contains. If audiences outnumber angles, that is your imbalance.
Check your exclusions are current, especially existing customers.
Confirm your geography matches where you actually work.
Then set up one broad ad set alongside your current setup, with the same creative and budget, and compare after two weeks on qualified leads.
The wider piece on judging lead cost covers the ceiling any of this should respect, instant forms against landing pages covers where the enquiry lands, and the order I work in to reduce lead cost puts targeting in its proper place near the bottom. For the specific ways budget disappears in a Meta account, the three places money leaks is the short version. To have your setup reviewed, book a teardown.