Modeled frequency management for anonymous inventory

You can use modeled frequency management in Display & Video 360 to increase your reach on anonymous inventory, including inventory served to Safari 12+, while still managing frequency. 

How modeled frequency management works

Modeled frequency management in Display & Video 360 uses machine learning to give you the ability to apply frequency caps to non-personalizable inventory in a privacy-safe way by using a model of frequency caps that's been trained using inventory that was eligible for measurement. Once your advertiser has been enabled for using modeled frequency management, all anonymous inventory sold by Google Ad Exchange will use modeled frequency management to give you the ability to control ad exposures while still being able to purchase anonymous inventory.

Modeled frequency management for anonymous inventory sold by third-party exchanges isn't currently supported.

What line items will use modeled frequency management?

Once your advertiser has been enabled to use modeled frequency management, any line items, insertion orders, or campaigns with frequency caps that buy anonymous inventory will use modeled frequency management when bidding on anonymous inventory, such as inventory from Safari 12+. If those same line items bid on inventory for which a third-party cookie is available, deterministic frequency capping will be used to achieve as close as possible to the desired frequency cap.

Note, as of fall 2019, the majority of anonymous inventory is associated with Safari 12+.

What if my advertiser doesn't want use modeled frequency management?

If you don't want to use modeled frequency management at all, either set your browser targeting to exclude Safari or remove your frequency caps. 

How can I purchase Safari 12+ inventory?

To purchase Safari 12+ inventory, you must either use modeled frequency management or not use frequency caps at all.  

How can I report on which impressions used modeled frequency management

In reporting, you can use the "Anonymous Inventory Modeling" dimension to see the number of impressions that were purchased that used modeled frequency management. Impressions that were marked as "Probabilistic" are impressions that were bought using modeled frequency management.

Was this helpful?
How can we improve it?