It’s been six months since the IETF 111 when Google engineers first announced their vision of a new iteration of FLoC built from site topics. Now, we finally have more details, with Google Chrome revealing its new Topics API system. The high-level, technical overview released by Google explains that through using this system, each website’s hostname will be categorised from an overall list of 350 topics.
Finally, Google has released an update for the advertising ecosystem, providing an explainer that can be analysed. However, there’s still more information needed before we can understand its full implications. In this article, I will explain Topics API, its possible uses, and the potential challenges still yet to be explained.
Topics API and Privacy
The concept of Ad Topics has been explored by many companies as a tool for reliable and private interest-based advertising. Meta, for example, conceived an Ad Topic Hints proposal that builds on user feedback to display ads and PAURAQUE’s proposal from NextRoll allows users to define interesting topics for personalisation. There are many challenges still to be overcome, for example, in the current shape of the proposal, it’s unclear yet whether publishers will have ultimate control over the topics their sites represent or whether an in-browser algorithm will assign topics for them.
Data privacy is at the centre of this proposal and is ultimately the perspective from which Topics API will be scrutinised most. A clear benefit gained from using generalised topics is the negation of third party data and the fact that users become anonymous within groups. When privacy is considered, there’s no question that this is a move in the right direction.
A positive component of this is the inclusion of an opt-in mechanism for websites rather than the widely and fairly criticised, original proposal of opt-out. This positively impacts websites and users. Websites gain the ability to assess the value of the Topics API system and users gain complete control, allowing them to opt-out at any time.
Effective ad targeting
For publishers and marketers, a big concern is how this will affect targeted advertising. In Topics API, users are assigned 5 topics within a set of 350, resulting in a number combinations which are theoretically larger than FLoC. As a result, there are several areas of advertising that could work efficiently.
In a traditional approach to internet advertising, brands should theoretically be able to reach new audiences using topics relating to users’ interests, taking advantage of the large interest-related groups available. To make this even more effective, this approach could be used alongside contextual targeting too. As the 350 topics are closed, contextual provides a greater level of precision when targeting while also giving further control over when to serve relevant ads. Both Topics API and Contextual targeting can complement each other using deep learning algorithms – Topics API can determine a user is interested in soda, for example, while contextual targeting can analyse keywords in the article. Together, it’s possible to build a detailed user profile while maintaining privacy.
Additionally, brands will be able to decipher their customers’ most popular topics, using that data to instigate an effective, new-user acquisition strategy using advertising. Again, these approaches can be completed while maintaining user privacy.
However, there is the issue of topic selection – any bias towards a small group of overpopulated topics will make advertising significantly more difficult, reducing the possibility of targeting granular interests. For example, anyone consuming news on huge social platforms such as Youtube and Facebook can be applied to topics, which is a huge proportion of the web population and introduces lack of effective utility. The Topics assignment algorithm will need to ensure that a user will get valuable information, even if the user spends the majority of time on social media. We still need more information on how this mechanism will work for these dominant websites.
But what about the competition?
Despite the general explanation of the new Topics API system, there are many questions that need answering before we can truly gauge its effectiveness. To start with, Topics API appears to have a skewed focus towards larger entities, rather than SMEs, bringing into question its market competitiveness. The reason is found in the mechanics of the system – the more websites a company appears on, the more likely it is to get topics to target, disproportionately favouring large social networks, multicategory publishers, supply-side platforms, and ad networks.
Added to this, each topic will be mapped to the website hostname rather than the full URL. This means that a larger website containing countless sub-topics, such as any major news publication, will only match topics linked to the overall website domain. If a news outlet has a smaller subcategory on tennis, for example, it may not be able to link to that keyword. Equally, a website dedicated to one topic may gain no useful information from its users at all. This poses a problem for publishers with topic groups potentially being too large.
It also relied heavily on a social effect, eg, the more websites that participate, the more valuable Topics information will be. Because there is less value for smaller publishers, then inevitably, only larger publishers will participate, resulting in the majority of users getting generalised topics such as ‘news’. The challenge is to provide value to both small and large publishers.
To conclude, this is vital news for the industry and is the beginning of what should be a constructive discussion on how we can successfully adapt to maintain user privacy, advertising efficacy and true competitiveness. Any new solution works best when implemented with the cooperation of the entire supply chain, and if the full testing period starts as soon as this quarter, we will hopefully gain answers to the industry’s most important questions, providing enough time for the industry to work together.
VP of Programmatic Ecosystem Growth & Innovation, RTB House
RTB House is a global company that provides state-of-the-art marketing technologies for top brands and agencies worldwide. Its proprietary ad buying engine is the first in the world to be powered entirely by Deep Learning algorithms, enabling advertisers to generate outstanding results and reach their goals at every stage of the funnel. Founded in 2012, the RTB House team comprises 750+ specialists in over 30 locations around the globe. It serves more than 2,000 campaigns for clients across the EMEA, APAC, and Americas regions.