When we start delivering our ad set, whether at the start of a campaign or after you edit it, we don't have all the data necessary to deliver it as stably as possible. In order to get that data, we have to show ads to different types of people to learn who is most likely to get you optimisation events. This process is called the "learning phase". Learn about editing the learning phase. Learn about editing during or after the learning phase.

• Once we have the data we need, your ad set should experience fewer performance fluctuations. At this point, you can make an informed decision about your ad set. If you're satisfied with your results, you can let it keep running or increase its budget. If you're unsatisfied, you can edit the ad set to try to improve its performance or pause it.
• During the learning phase, you can expect more performance fluctuations than usual. We recommend against making significant edits to your ad set during the learning phase. Doing so may cause the learning phase to reset before generating any meaningful information.
• To complete the learning phase, your ad set needs to get about 50 optimisation events after starting to run or since the last significant edit.

If your conversion window is 7-day click, the 50 or so conversions you need all have to happen within seven days of the click that led to them. This means that ad sets with longer conversion windows may need more time to complete the learning phase, as they're probably promoting products or services that have longer consideration times.

* Nguồn: Facebook