Enabling and honoring frequency capping is one of the most fundamental features that a bidder provides. It is also one of the most common optimization lever that a campaign manager uses to optimize their campaigns.
By enabling frequency capping, the campaign manager limits the maximum number of times the user can see an an ad of a specific brand. To make this work, the system needs to identify the user and the most common and deterministic way of identifying the user in mobile app ecosystem is via device identifiers- GAID (android) and IDFA (iOS). However, not all bid requests that we receive have these identifiers. We are happy to announce that we have released a probabilistic model for frequency capping which now enables advertisers to set frequency capping even for those bid requests where deterministic user identifiers are not present
Today, ~10% of all traffic that we receive do not have device id. The ratio changes based on different inventory combinations. For instance, ~21% of US iOS traffic does not have device identifier while the same ratio increases to 32% for EU iOS traffic. You can see this data via our Inventory Analytics platform. There are multiple reasons why we do not get device identifier in the bid request. Users might have activated LAT (Limit Ad Tracking) functionality in their phones or because of regulations like GDPR/CCPA , we may not get user identifier data in the bid request . With the impending release of IDFA related opt-in changes in iOS 14, the percentage of such requests will continue to increase.
A lot of savvy campaign managers are already using the platform to target such users for whom we do not get device id in the bid request by leveraging "Missing Device Identifier Targeting" feature.
With the release of probabilistic model for frequency capping, the advertisers can now enjoy the benefits of frequency capping even for those bid requests where deterministic user identifier is absent. This model takes multiple parameters as input like User agent, IP address, device characteristics of the user (device model, screen size, os, osversion, manufacturer and more). In the internal A/B tests, for the campaigns which were targeting bid requests with "missing" device identifier, we saw considerable performance improvement in terms of different performance metrics like CTR, CPI, IPM etc when the campaign was using the probabilistic model vs not using the model at all.
The feature is enabled by default for any campaign which is targeting bid requests without device id.
Should you have more questions or feedback about the release, please do reach out to your Programmatic success specialist