Not immediately. Since some volatility is expected during a learning phase reset, most sources recommend waiting several days for delivery to stabilize before making further changes, unless the CPA is running at several times your target.

A TOF strategy is a separate campaign layer focused on finding new prospects, often optimized for a higher-funnel event like landing page views or add-to-cart, rather than purchases. It expands the pool of people available to the algorithm, which reduces the pressure that causes CPA to rise as you scale.

It happens when your budget outgrows the size of your addressable audience. Meta has already spent on your best, cheapest-to-convert prospects, so additional budget goes toward reaching colder, lower-intent users, which raises your average CPA.

Not always. A learning phase reset is usually temporary and stabilizes within days. Audience saturation and auction pressure, the other two causes, are structural and will keep pushing CPA up over time unless you expand the pool of people the algorithm can reach.

An ad set typically needs to accumulate around 50 optimization events within a 7-day window to exit learning. For accounts with lower conversion volume, this can take longer and may result in a “Learning Limited” status.

Industry sources consistently point to a 20% threshold. Increases under that amount, made in increments every 3-5 days, are less likely to trigger a full reset. This is a widely used guideline, not an official published rule from Meta.

A budget increase can trigger a learning phase reset if the change is roughly 20% or more, and it also forces Meta to reach further into your audience, past your best-converting prospects, and bid more aggressively to win the extra impressions. Together these push CPA up, at least temporarily.A budget increase can trigger a learning phase reset if the change is roughly 20% or more, and it also forces Meta to reach further into your audience, past your best-converting prospects, and bid more aggressively to win the extra impressions. Together these push CPA up, at least temporarily.

There is no single review frequency appropriate for every account.

A high-spend account can accumulate enough data to reveal problems quickly.

A small campaign may need considerably longer before individual segments have enough data to support useful decisions.

Instead of changing campaigns according to an arbitrary schedule, monitor:

• spend velocity;
• CPC movement;
• conversion volume;
• CPA;
• conversion value;
• search-term quality;
• major tracking changes;
• and material performance deviations.

The goal is to respond quickly to meaningful problems without reacting to normal day-to-day variation.

Broad match shouldn’t be evaluated solely on whether its CPC is higher or lower. Evaluate the search terms it discovers and whether those searches generate incremental conversions, qualified leads or conversion value at acceptable economics.

Broad match gives Google’s matching system more flexibility than phrase or exact match.

That can be valuable when:

conversion tracking is reliable;
the campaign has meaningful conversion signals;
irrelevant searches are monitored;
negative keywords are maintained;
and bidding is aligned with the desired outcome.

If those foundations aren’t in place, broader matching can make existing weaknesses more expensive.

The answer therefore isn’t universally:

“Use broad match.”

or:

“Avoid broad match.”

It is:

Test whether broader matching improves your specific acquisition economics.