Free tool
Meta Learning Phase Calculator
A Meta ad set exits the learning phase when it gets enough results. From your cost per result, the tool works out the budget that funds those results each week, and how many your current budget funds.
The only figure in Meta’s developer documentation is the one for instant games ads: approximately 50 optimization events per week for each ad set. Meta’s API also returns, for each ad set, its learning status and its own conversions threshold.
Your Ad Sets Stay Stuck In Learning?
I restructure your Meta campaigns so each ad set gets enough results to learn.
20 minutes · No commitment · 100% confidential
How the budget to exit learning is calculated
Weekly budget = Target events × Cost per result Daily budget needed = Weekly budget ÷ 7 Events per week = Daily budget × 7 ÷ Cost per result
At 20 € per result, 50 events a week cost 1,000 €, or 142.86 € a day. With 100 € a day, the ad set funds 35 events a week, 70% of the threshold.
Meta’s API describes three statuses: still learning, exited the learning phase, and “not generating enough results to exit the learning phase”. It counts conversions since the last significant edit, which restarts learning: an ad set you keep editing never gets out.
The calculation keeps your current cost per result. When the budget goes up, that cost often goes up too: the marginal ROAS calculator measures what the added euros really cost. If the budget needed is out of reach, merge ad sets, or optimize for an event more frequent than the purchase.
What the calculation lets you do
- Give each ad set a budget it can learn with, instead of splitting the budget into parts that are too small.
- Decide how many ad sets your budget can support: a total budget divided by the budget needed gives their maximum number.
Frequently asked questions
How many events does it take to exit the Meta learning phase?
Meta’s developer documentation gives approximately 50 optimization events per week for each ad set, for instant games ads. Its API also returns a threshold specific to each ad set: that is why the threshold can be changed here.
How do you calculate the budget to exit learning?
Multiply the target number of events by your cost per result, then divide by 7 for a daily budget. 50 × 20 ÷ 7 = 142.86 € a day.
What restarts the learning phase?
A significant edit to the ad set. Meta’s API records the date of the last one, and counts conversions again from there.


