What Meta says
In Meta's own words: “Your performance goal is the desired outcome that our system bids on in the ad auction.” (Meta help center)
Why the choice matters more than it looks
Every delivery decision is made by pattern-matching against people who already triggered your optimization event. Choose a frequent event and the algorithm accumulates real examples quickly. Choose the rarest event in your funnel and it is guessing, because it has barely seen anyone do the thing you asked for.
Frequent, but still meaningful
The trap on the other side is picking something so shallow that anyone triggers it. Optimizing toward a page view teaches the system to find browsers. Pick the shallowest event that still indicates genuine intent, then move deeper as volume allows. On Meta this interacts directly with learning limited; on Google, the equivalent question is whether your bid strategy is pointed at the conversion goal at all. We work through the Meta case in don't optimize a new Meta campaign for purchases.
A worked example
Say an ad set spends $700 a week and a purchase costs $70: that is about 10 optimization events a week, nowhere near the roughly 50 Meta looks for, so it sits in learning. Switch the optimization event to add to cart at $10 each and the same $700 buys about 70 events a week, enough for delivery to stabilize. The trade is that the system now learns who adds to cart, which is why we move deeper again once volume allows. Click-to-WhatsApp campaigns are a common trap here: in a Leads campaign that sends people to WhatsApp, conversations are the only goal on offer, and a conversation isn't a lead: WhatsApp ads: conversations aren't leads.
Common questions
What is an optimization event?
The action you tell the ad platform to optimize delivery toward, such as a lead, an add to cart, or a purchase. The algorithm uses people who triggered it as the pattern for who else to show ads to.
How do I choose the right optimization event?
Pick the shallowest event that still means something commercially and happens often enough for the platform to learn from. Too rare and delivery never stabilizes; too shallow and you optimize toward people who were never going to buy.
Does Meta need 50 optimization events a week to exit the learning phase?
Roughly. Meta says ad sets usually exit learning after about 50 results in the week after the last significant edit, and flags an ad set as learning limited when it is unlikely to receive about 50 optimization events in that week. Shops ads are the exception: 17 website purchases and 5 through Meta.
Is an optimization event the same as a performance goal?
They describe the same choice. In Ads Manager you pick a performance goal, which Meta calls the desired outcome its system bids on in the auction; Meta's learning phase docs count each such action as an optimization event.
Does changing the optimization event reset the learning phase?
Yes. Any change to the optimization event is on Meta's list of significant edits, so the ad set reenters the learning phase.