"Learning Mode” occurs at the beginning of any new performance campaign launch. It is during this time period that the algorithm buys a wide variety of media with the ultimate goal of “learning" which placements produce the most efficient conversions.
Why is there a learning mode?
Learning mode puts guard rails in place for a performance campaign as it builds signal. This prevents a campaign from wildly overspending while not hitting your goal.
How does learning mode bidding work?
Learning mode is largely based on the information obtained via a conversion pixel placed on an advertiser’s page. Every time a pixel successfully fires and registers a conversion, the pixel records a variety of variables (such as publisher, template, ad slot, gender of person, age, and more.)
The more conversions the algorithm obtains, the more accurately it can predict where the next one will come from. This, in essence, sets the parameters for what inventory the algorithm will bid on, and how much it will bid.
What happens next?
Learning mode ceases when the linked pixel is able to register 10 conversions. However, if either of the following happens before it registers these 10 conversions, the campaign will shutdown:
1) The algorithm spends 10x the inputed CPA goal, or
2) The algorithm spends 10% total inputed budget.
These “fail-safes” are put in place to ensure that spending will stop if a campaign is not bidding efficiently and hitting goal.
If a campaign does shutdown, a relaunch with changes to the goals is needed