Learning period
The learning period is the time an automated bid strategy or ad set needs after launch or a major change to calibrate. Results are less stable then.
In the platformAutomation
Learning
The learning period is the time an automated bid strategy or ad set needs after launch or a major change to calibrate. Results are less stable then.
While it lasts, delivery and cost per result swing more than usual, and judging a change on those days usually leads to the wrong conclusion. A new bid strategy, a large budget or target change, new conversion actions, or, on Meta and TikTok, big edits to targeting or creative can each restart the clock.
Where you'll see it
In Google Ads, the Bid strategy status column shows "Learning" for Smart Bidding campaigns, with a short explanation of the cause. In Meta Ads Manager, the Delivery column shows "Learning" for ad sets, and "Learning limited" when an ad set is unlikely to get enough optimization events; Meta's guidance is roughly 50 within 7 days of the last significant edit. TikTok shows a similar status on ad groups, with comparable guidance.
Worked exampleTarget CPA from $40 to $32
- Days 1–5 (learning): CPA ranges from $28 to $55.
- Reacting on day 2 at $55 and raising the target again restarts learning.
Wait out the period, then compare at least a full week with the week before.
Do
- Change a target in small steps
- Space out changes so the campaign can leave learning
Don't
- Make several changes a week
- Run so many Meta ad sets that none gets enough events
- Compare learning days with a stable baseline