Interest Groups and Custom Audiences – What Are They?

For the time being cookies are an integral part of online marketing. They allow for user IDs to be matched with tons of data that is useful for ads personalization. These IDs can be shared between different entities. For example, when an ad is generated for a user, the Supply Side Platform (SSP) sends a bid request with the user's ID and some other information, and the Demand Side Platform (DSP) takes that ID, and, in turn, combines it with further additional data. The idea is simple and effective in terms of bringing great results for advertisers, and convenience for users, but over time privacy concerns have grown.

The reason for these concerns is clear. Information contained in cookies is hard to access and illegible for a user. As a result a typical user can never be sure who keeps their data, what kind of data it is, and with whom it is shared. Even though the law requires every entity to provide the possibility to opt-out, the process is often complicated and users may not fully understand which data they have opted-out from and which data is still collected.

Interest Groups – the foundations

The industry is preparing itself for a future without cookies. Soon, user IDs will no longer be permitted for the purpose of ad personalization and building cross-site audiences. Rather, information about a given group itself may instead be shared. The idea of Interest Groups in Protected Audience API provides the required user anonymization and simultaneously addresses the need for targeting users who may be interested in specific products or services.

On the basis of a user's activity on a given website, they are assigned to specific Interest Groups. Importantly, the activity informing Interest Group assignment has to come from one website only―it cannot be shared between domains. The information about assigned groups is kept on the user’s device, in the browser, separated from the rest of the online advertising ecosystem. Additionally, the information is under the control of the user who can at any time view the Interest Groups they are added to and even remove themselves from the ones they don’t like. A user assigned to a specific Interest Group can be targeted not as a specific identifiable person, but as an anonymous member of a group.

Typical application of Interest Groups

Within Protected Audience API, Interest Groups are typically created by the delegated demand-side adtech vendor. The creator is always indicated in the Interest Group as the ‘owner’. This means, when the auction is performed, the DSP can bid only for groups it has created. 

In the simplest scenario, the DSP creates Interest Groups on the advertiser’s website on the basis of user behavior saved in a partitioned, first-party profile. As an example, it may include information about products visited by the user. For example, if a user visits a shoe shop and browses several running shoes, they might get added to Interest Groups like ‘shoes’, ‘sport’, ‘sport-shoes’ or ‘fitness’. However, the name of the Interest Group doesn’t have to clearly indicate the interest. Perhaps the DSP’s algorithm finds more nuanced similarities between users than just visited products.

The publisher-based approach has enormous untapped potential

Less obvious is that Interest Groups can also be created on the publisher’s website. Typically this approach would also require a specialized adtech vendor capable of the creation and activation of such groups. After agreeing on the partnership, and implementing an adtech’s code on the publisher's website the magic happens. 

Based on a publisher’s first party data, such as interactions between a user and a website, and data accessible to the adtech vendor, like TopicsAPI information, the Interest Groups can be created. The owner can of course bid for members of these groups, not only on the inventory that belongs to the publisher whose data was used to create groups, but also on other publishers’ websites. This creates a whole new ecosystem where publishers’ first-party data can be used to inform a DSPs bids outside of their inventory in a privacy-preserving, transparent, and controllable way that benefits the original publishers. This is a huge improvement over today’s status quo where data brokers simply steal publisher data by tying it to cross-site user IDs. 

How it will work in practice

If a user browses a news page VeryNiceNewsPage.com and reads an article about being fit, they might get added to ‘fitness’, ‘lifestyle’, and ‘sport’ Interest Groups. These groups will remain on their browser for the next 30 days (it can be prolonged if a user gets added to the same group again) or the user can manually remove a group from their browser. As long as they are assigned to the group, the DSP is able to take part in an auction and bid for showing the user a relevant ad. There will be a unique bid request for each Interest Group. Otherwise, if the bid request includes information about all groups assigned to the user, this  more easily allows for user reidentification. 

For example, based on the group ‘sport’ the user might get an ad with running shoes or sportswear if this bid wins. This benefits both the publisher selling their inventory and the publisher the data originates from. In many cases it will be the very same publisher.

Android – the kingdom of custom audiences

The idea is quite similar when it comes to Android devices, but the mechanism is just a bit different here. Protected Audience API allows for the grouping of users on the basis of interactions within a specific app. That can include, for example, the user’s in-game engagement like finishing the intro guide, reaching a specific level, or buying something in the game for the first time. The Custom Audience lifetime is still under discussion, but the value won’t be fixed, as for the Interest Groups, just predefined with the possibility to set a custom lifetime. 

The mechanism of user access to Custom Audiences will also be a little different. Users will be able to access the list of apps that created any custom audience on their device. If a user doesn’t like the result, they’ll be able to remove apps from the list together with their set of custom audiences. Such a move also blocks the app from creating new custom audiences for this device. Generally, despite some technical differences, it gives very similar possibilities from the user’s perspective. The applications for adtech vendors will also be similar―it will be possible to create and target audiences in a similar way, including the possibility to input buyer-side code into a publisher’s app to create Custom Audiences that can be reached on the same or other apps.

The future of grouping

Despite the idea of segmenting users not being new to the market, the approach of Interest Groups and Custom Audiences will definitely be key in the cookieless world. It allows the leveraging of many types of first-party party data from different sources; from the advertiser’s or publisher’s side, depending on the need. Reaching users will be possible even outside the original context, where the segment was added to allow publishers an additional revenue stream for their first-party data. Above all, it’s a solution that satisfies the key requirement of user privacy.

‍