Bye-bye FloC, Hello Topics API

Before we dive into Google’s new Topics API, let’s start with a crash course on what’s happened with third-party cookies over the last two years.

January 2020: Google announced that they will deprecate third-party cookies by 2022. They launched the Privacy Sandbox initiative to curate proposals on viable alternatives.

March 2021: Enter Federated Learning of Cohorts (FLoC). Google’s alternative to third-party cookies. Except, it was riddled with issues noted in feedback received by publishers, advertisers, and the ad tech industry in general.

June 2021: Everyone sighed with collective relief as Google announced that they were delaying the deprecation of third-party cookies until 2023. FloC was still in testing, but Google needed more time to correct its course.

January 2022: We did say crash course. As per their recent announcement, Google has now decided to replace FloC with the Topics API.

This new API proposes that the web browser identifies certain topics that reflect an accurate depiction of a user’s interest. The identification will be based on the users’ browsing history on a weekly basis with the help of participating websites.

Let’s unpack this.

What is the Topics API?

The Topics API is born out of feedback received on the FloC test trials. As a refresher, FloC aimed to group users into interest-based cohorts gleaned from their browsing history.

With the Topics API, the goal is to personalize advertising through a set number of content topics that are regularly updated by the browser. What does that mean?

Similar to FloC, the Topics API leverages in-browser machine learning. Unlike FloC, it gives the browser the ability to select three topics (from up to five) to map a user’s main interests.

This process is conducted over three weeks across specific, participating websites that choose to call this API. The Topics API has three main goals:

1. Map the website hostname to interests. For example, a visit to finance.example.com may be classified as a user’s interest in the Finance topic.
2. Calculate the top five topics based on the user’s browsing history.
3. Offer a JavaScript API to websites that will help the selection of personalized ads.

Shared topics are linked to a website and will last for three weeks. After that, the topics are deleted and replaced with fresh topics. The API offers multiple topics so they can be combined. For example, someone interested in their financial success who is also interested in gardening.

With this proposal, there will be less information available to target users, especially since the new data will come with an expiry date. However, “[t]hat provides a more private alternative than tracking to help deliver relevant ads,” Ben Galbraith, Senior Director, Product at Google.

How does the Topics API work?

Publishers will need to include the API code on their website in order to action any API calls. It may look like this,

// Get the array of top topics for this user.
const topics = await document.browsingTopics();

// Request an ad creative.
const response = await fetch('https://ads.example/get-creative', {
 method: 'POST',
 headers: {
   'Content-Type': 'application/json',
 },
 body: JSON.stringify(topics)
})

// Get the JSON from the response.
const creative = await response.json();

// Display ad.

Without this code in place, the Topics API will not run.

Let’s assume a publisher adds the API code to their website.

When a user visits their website, the Topics API selects three interest topics for the user to share with that publisher and its advertising partners. That’s one topic from each week in the three-week timeframe. Up to five topics can be determined for a user based on their browsing activity. The topics are then kept for three weeks before they are deleted and the process begins once again.

How about an example?

A user visits websites associated with finance, gaming, sports, pets, and real estate over the course of three weeks. The Topics API will choose the top three topics of interest to share with advertising partners. After three weeks, a new set of topics is chosen and the process repeats. Chrome has created a visual representation to map the process.

What caused the change?

One of the biggest concerns with FLoC was its potential ability to create cohorts that would allow the technical conditions for discrimination against users based on their browsing interests. Google hopes to rectify this concern through the Topics API.

We’ve summarized the main differences below:

FloC	Topics API
Uses interest-based tracking to bucket users into cohorts	Uses interest-based tracking to bucket users by interest topics
Analyzes a user’s browsing habits and groups them into a cohort, based on the previous week	Analyzes a user’s browsing history to determine their top three interests over the course of three weeks
Shares cohort ID with advertisers	Shares top three interest topics for a user with advertisers for up to three weeks before the data is erased

How does the Topics API address FloC concerns?

Reduces fingerprinting

The current Topics taxonomy lists approximately 350 topics. Each topic should have a fairly large user base.

Let’s say a user visits finance.example.com every week. If the publisher who owns finance.example.com participates with the Topics API, the code running on its site can only learn one new topic per week, at most. Different websites will see unique topics for the same user for the same timeframe.

There’s a 20% chance that the topic generated for a user from one site will match the topic generated for them on a different website, making it more difficult to identify unique users. Additionally, Topics are updated weekly, thus limiting the rate at which information is shared.

Sensitive topics

Earlier, we mentioned how FloC was Google’s alternative to third-party cookies. Specifically, an alternative that would not rely on cross-site tracking. But it was met with many challenges as noted by the ad tech industry.

Criteo found that it was still possible for FloC IDs to track users across websites by observing browsing habits and cohort assignments over time. Meaning, users could still be grouped into “sensitive categories,” such as their ethnicity.

In order to rectify this, the current list of Topics was manually created in order to avoid including sensitive attributes and categories.

Transparency for users and websites

By making the aforementioned list public, Chrome hopes that users will be able to learn more about the topics that can be associated with them as suggested by their browser. The goal is to make the information about the Topics API available for users to access at any point, most likely through Chrome settings.

Users will also have the power to opt-out of the Topics API should they decide to not share that information with publishers and advertisers.

Additionally, only websites that use the Topics API code will be included in the topic frequency calculations. It will not be automatically enabled for all websites accessed through Chrome.

In summary

Google aims to resolve issues associated with FloC through the Topics API. Specifically, by identifying topics based on website hostnames and browsing activity. The goal is to reduce fingerprinting and provide a more transparent ad personalization that respects user privacy.

While this update is fairly new, it is still just a proposal. A proposal that will need to be accepted by different browsers like Firefox, Safari, etc., as well as advertisers and publishers.

Additionally, the Topics API only runs through participating websites that use its API code. Meaning, not every publisher may choose to enable and use the Topics API.

While the Topics API is meant to replace FloC, Google has not changed the 2023 deadline for the deprecation of third-party cookies, which includes the testing and transition periods that can be accessed through the Privacy Sandbox timeline.

You can find a more technical description of these changes and how the Topics API will work through Chrome Developers.