My campaign never seems to get out of learning. Is my optimization event too rare?

Learning is a per-ad-set problem, not a per-account one

The signal budget is spent at the ad set level. Splitting a budget across six ad sets that each target a slightly different slice of the same metro divides your event volume by six and leaves every one of them starved. The account looks busy and nothing converges.

This is why consolidation usually beats clever segmentation for local service businesses. You rarely have enough demand in one service area to feed many parallel learners. One well-fed ad set generally outperforms six hungry ones, even if the six look more sophisticated in a slide deck.

Choosing the right event on the value-versus-volume curve

Events sit on a spectrum. Page view is abundant and nearly meaningless. Completed job is meaningful and rare. Between them sit form submission, qualified call, appointment booked. Your job is to pick the most meaningful event that still occurs often enough to learn from.

A useful test: count how many of that event your worst-performing ad set produced last week. If the answer is small enough to be affected by one dispatcher having a bad day, the event is too rare to optimize on and you should move one step up the funnel.

What relearning costs you

• Edits reset it. Changing the budget substantially, the optimization event, the audience or the creative set can push an ad set back into learning. Constant tinkering is expensive in a way the interface does not show you.
• Erratic results invite more tinkering. The loop is self-reinforcing: unstable performance prompts changes, changes prevent stability. Setting a review cadence and holding to it is a real intervention.
• Small budgets amplify it. The less you spend, the longer any given event count takes to accumulate, so low-budget accounts should optimize on more frequent events than high-budget ones.

Reporting on outcomes while optimizing on proxies

The resolution most operators need: optimize on an event frequent enough to learn from, and report on the outcome you actually care about. Those do not have to be the same event. You can bid toward qualified leads while judging campaigns on booked revenue, as long as the two are connected in your own reporting.

That separation is only possible when the leads are matched to jobs outside the ad platform, which is the point of connecting Meta to your operational systems. Judge with revenue data; steer with whatever signal is dense enough to steer with.