There is one setting in a Meta account that decides more than all the targeting put together, and most people set it once and never look at it again. The Meta ads optimization event tells the algorithm which action to go and find more of, and it will do exactly what you asked with impressive efficiency.
Ask for page views and you will get page views. Cheap ones, in volume, from people who had no intention of enquiring.
This is for you if your Meta leads look fine on the dashboard and disappointing in the inbox.
What the Meta ads optimization event actually controls
It is the instruction, not a preference.
Meta builds a model of who is likely to perform the action you named, then spends your budget finding those people. Everything else, the audience, the placements, the bidding, works in service of that one target.
So the event is the sharpest lever in the account. Change the audience and you adjust who is eligible. Change the event and you change what the system is trying to achieve.
That is why an account with excellent creative and the wrong event underperforms a mediocre account with the right one.
How the wrong Meta ads optimization event hides
It hides because it produces good-looking numbers.
Optimise for landing page views and you will get them cheaply, because loading a page is easy and lots of people do it. Cost per result falls. Volume rises. Every chart points the right way.
The damage appears somewhere the ad dashboard cannot see: enquiries that do not answer the phone, or a sales team quietly deciding the ads are not worth following up.
Most Meta accounts I audit are optimising on the wrong event, and it is usually because someone configured it once and nobody looked again. It is not carelessness. There is simply nothing in the interface that flags it, and the reports it produces look like success.
How this happens in practice
Three routes, all common.
The account launched before conversion tracking was properly set up, so an earlier event was chosen as a stopgap and then became permanent.
Someone duplicated a campaign to make a new one, and the duplicate carried the old setting. Duplication is how a wrong event spreads, because it copies what was there rather than what you intended.
Or the website changed, an event stopped firing, and rather than fix the tracking someone switched optimisation to an event that still worked.
That third one is worth naming clearly. If your Lead event stopped firing, the fix is the tracking, not a different target. This is where running the Pixel and the Conversions API together earns its keep, because a server-side signal survives the things that break a browser one.
Checking your Meta ads optimization event properly
Open each ad set, not the campaign, and read the performance goal and conversion event fields.
Check every ad set individually. The setting lives at ad set level, so one can differ from the rest after a duplication, and that is exactly the one causing your inconsistent results.
While you are there, confirm the event you named is actually firing. An event that is selected but never received leaves the system optimising blind, which produces erratic delivery and costs that jump around without explanation.
What happens when you fix it
Expect the headline figure to get worse.
If you move from page views to leads, Meta stops chasing the cheap, easy action and starts looking for people who submit enquiries. There are fewer of those and they cost more to reach. Cost per result rises, sometimes sharply.
That is the trade you want. Fewer, better, more expensive enquiries beat a pile of cheap ones, and the only way to see the improvement is to judge on qualified leads or closed business rather than on the dashboard.
Warn whoever reads the report before you make the change. Otherwise you will make the account better on Monday and spend Tuesday defending it. Say the number will rise, say why, and agree in advance what you will judge the change on. That conversation takes five minutes and saves a fortnight of argument.
When a lighter Meta ads optimization event is correct
Volume. If an ad set cannot gather roughly 50 conversions a week, the system never leaves its learning phase and delivery stays unstable.
In that situation, optimising on a rarer event genuinely does not work, and stepping back to something more frequent is the right call.
Try consolidating budgets first. Several thin ad sets chasing the same goal usually do better merged into one well-fed one. Only step back to a lighter event if consolidation is not enough, and treat it as temporary rather than a decision.
Why this costs more than it used to
Wasted spend used to be annoying. It is now expensive.
Meta reported average price per ad up 12% year over year for the quarter ending 30 June 2026, with impressions up 14%. Inventory costs more than it did.
So an account chasing the wrong action is buying the wrong people at a higher price. The error has not changed. The bill has.
There is a compounding effect too. Every conversion you feed the system trains its model. Feed it page views for a year and it gets very good at finding people who load pages. That learning is not neutral. It has to be unlearned, which is why fixing this takes a couple of weeks to settle rather than working overnight.
Set it right early and the model compounds in your favour instead.
One more reason to check today rather than next quarter. The longer it runs, the more history you are asking the system to unlearn. A month of wrong signal is a nuisance. A year of it is a rebuild.
What to change this week
Four steps.
Open every ad set and write down its optimisation event, without assuming any of them.
Confirm each named event is genuinely firing, not just selected.
If you find the wrong one, tell whoever reads the report what will happen to the number before you change it.
Then judge the change after two weeks on qualified enquiries rather than on cost per lead.
The wider piece on judging lead cost explains why the headline figure alone misleads, the order I use to reduce it puts this near the top for good reason, and a weekly rhythm for creative testing covers the other main lever. If you want someone to check this properly, Meta lead generation is my main work and a teardown starts here.