The How, What and Why of Topics API

On the 25th of January, Google announced the latest initiative within their Privacy Sandbox, “The Topics API”, as the new way to activate audience-led advertising in a future devoid of third party cookies. Topics won’t rely on tracking the sites a user visits, but will use their own browsing history to supplement contextual information to feed data  to publishers and ad exchanges about topics the user is currently interested in.

How does it work?

For the developers out there, Google has laid this out in detail, however in this blog,  I’ll cover the top line.

The browser will infer a handful of interest-based categories, or topics, based on recent sites that a user has browsed, to help sites serve relevant ads.

The Topics API labels each website with a high-level topic. Once per week, based on (preliminarily) the user’s previous three weeks of browsing history, a selection of Topics they are theoretically most interested in, as well as one random topic, will be selected as the user’s topics for the coming week. One new topic per week will be shared with the sites the user visits.

The initial selection of Topics (found here) bears a very strong resemblance to the currently targetable Google Affinity audiences and should allow DSP targeting to feel very familiar. There’s also the intention for the list of Topics to develop over time, increasing the level of granularity available to traders significantly.

Privacy: FLoC this is not

Google has taken quite a few steps with Topics API, and although a lot of the details are yet to be ironed out, it seems initially that there are significant strides towards an approach that will fit within the privacy sandbox in comparison to FLoC.

Firstly, all of the data processing will happen within the user’s browser rather than on a third party server. Consumers will have the option to remove specific topics from their selection, or opt-out of the entire process. Clearing one’s browsing history or using incognito mode will cause the topics returned to be empty when called, giving users a great deal of control over when and if they want to participate. 

To this end, the taxonomy of topics available will be human friendly, meaning it will be as easy as possible for anyone to see the topics ascribed to them, as well as the list of topics being human curated to prevent the inclusion of sensitive topics. By using a limited number of topics each week, as well as the inclusion of a random topic, the intent is to make it harder to fingerprint and identify individual users.

It’s also important to note that only third party callers present on a publishers site will be able to call topics gained by a user from that site. This will prevent third parties from collecting more data than they would have previously been able to glean from third party cookies.

There’s still a long way to go

As we examine the challenges we expect Topics to face on the road to widespread adoption and beyond, we should note that Google so far has come out with open arms and open questions, acknowledging that a lot of details are still up for discussion and there are answers we are yet to find.

The first of which (assuming all technical aspects are worked out), will be having enough publishers, data partners and browsers get involved to allow for this to have an effect on targeting across a large proportion of the web. I can see incentives for publishers as it may make advertising on their site more attractive, and data partners have everything to gain from collecting more data. Browsers may be the group that has the least to gain, but as Chrome is the browser of choice for the majority of internet users this may be more of a stumbling block than a brick wall, on desktop at least.

Next will be the users themselves. Individuals will have the ability to opt out of this entire process and there is certainly enough public sentiment against corporations handling their data that many may choose to do so. We’ll be relying on a significant rebrand or just sheer inertia to maintain a significant proportion of internet users willingly taking part, or more likely simply not actively removing themselves.

Next we of course have the effect on programmatic targeting itself. Each hostname will only be assigned a maximum of three topics, and with topics only updating once per week, traders won’t have the ability to target the most up-to-date information with the granularity to which we have become accustomed.

Also, it’s important to note that in its current form, Topics does nothing to enable future remarketing efforts. We’ll be relying on the implementation of other initiatives to allow for this lower funnel tactic moving forward.


At the end of the day, do we see Topics as a direct replacement for third party cookies? 

No, absolutely not.

But is that something we would even want? Google so far has been incredibly transparent with exactly how they intend Topics to work, their desire to eventually have certain aspects managed externally, and their request for community input. User choice, visibility and privacy seem to be at the forefront of the proposal, and we are cautiously optimistic about the future of Topics. The broadness of the targeting capabilities will increase the need for clever planning and smarter creative execution, and we do not see this as a be-all and end-all solution. We see this as a tool we can use to enhance contextual targeting and combine with other developing technologies to maintain efficiency and protect privacy in the ever so quickly approaching cookie-less future.

If you have any more questions about Topics API or how we can help, get in touch.

by Harry Owens
10 February 2022



Related posts