Topics API latest updates

# Chrome’s commitment to Topics

January 24, 2023

Following the W3C Technical Architecture Group’s initial design review of Topics, we’ve received some questions from ecosystem stakeholders about what this means for the Topics API.

As we noted on Twitter, we want to affirm our commitment to make the Topics API available in Chrome Stable this year and continue the public development process. The Topics API offers significant privacy improvements over third-party cookies, limits cross-site tracking, and is an important building block for a more private web. It will provide publishers and advertisers with an important signal to inform interest-based advertising once third-party cookies are no longer available.

Launching Topics in Chrome will provide the web community with the opportunity to observe how Topics performs in the real world, as we continue working toward the longer-term goal of browser interoperability. We’ve seen a tremendous amount of engagement across the ecosystem to design and begin testing Topics, and we are excited to improve the API in 2023 and beyond.

# Insights from early Topics testers

January 9, 2023

Topics testing in Chrome is still at an early stage, focused on validating the technical stability and core functionality of the API. However, we are encouraged to see some early testers exploring aspects of the business logic behind Topics and sharing their insights publicly. For example, Xandr looked at how the Topics classifier compares to Xandr's own method of categorizing websites, and Criteo evaluated the ability of Topics, as a standalone signal, to predict users' future interaction with advertiser websites. While isolated analysis won't meaningfully predict the real-world performance of the Privacy Sandbox technologies, it can prompt constructive dialog and surface areas for improvement. As more holistic utility testing becomes possible in 2023, we look forward to deeper industry engagement to optimize Topics and develop best practices for its use in digital advertising products. If you are testing Topics we encourage you to share your plans and insights on the Topics API Tester List.

# Ecosystem feedback on Topics for Q3 2022

October 27, 2022

As part of our commitments to the CMA, Chrome publishes quarterly feedback reports on the Privacy Sandbox proposals, summarizing feedback received from the various sources including GitHub Issues, the Privacy Sandbox feedback form, meetings with industry stakeholders, and web standards forums. The 2022 Q3 report includes Topics feedback themes such as the accuracy of the Topics system for inferring topics of interest from hostnames, the granularity of the Topics Taxonomy and the usefulness of Topics for various types and sizes of websites. (Past reports: 2022 Q2 | 2022 Q1) Here's some general guidance on providing feedback; scroll down this page to the "Help Improve Topics" section for some specific areas where we are seeking ecosystem input.

# Topics origin trial increasing to 5% of Chrome users

October 26, 2022

Chrome has started to increase traffic for the Privacy Sandbox Relevance and Measurement origin trial, including Topics, from 1% of Chrome Stable traffic to 5%. The trial has been available in Chrome Stable since August, and feedback from early testers has helped improve API stability so that we can now expand the trial population to continue functional testing through 2022 and prepare for utility testing in 2023. Stay tuned for more detailed utility testing guidance to help testers evaluate the Topics API for their use cases. If you'd like to receive notifications about origin trial progress and other developer updates, please join the Topics API Announcements email group.

# Topics tester page launched on GitHub

October 11, 2022

To help consolidate information about Topics testing, we've created a Tester List page on GitHub where Topics testers can identify themselves and link to their learnings. This list is voluntary and self-reported, so we don't expect it will be complete or representative of all testing activity—but we hope it will be a useful hub for testers who are willing to share their insights with the community and inspire others to get involved. If you are testing Topics or making plans to test, please add your organization to the list. You'll find detailed instructions on the page.

# Help improve Topics

The Privacy Sandbox team welcomes all feedback regarding the design, implementation and effectiveness of the Topics API. You can join the discussion and raise questions in the issues for the Topics proposal on GitHub. You can also provide feedback via the Privacy Sandbox feedback form.

Here are some specific areas where the Chrome team is seeking input from testers and other stakeholders.

# Topics taxonomy

The initial taxonomy for the web version of Topics includes around 350 topics across categories such as "Arts & Entertainment," "Home & Garden," and "Travel & Transportation." Although the list is human-curated to exclude explicitly sensitive topics, we acknowledge that some topics may have unintended correlations to sensitive topics. The eventual goal is for the taxonomy to be sourced from an external party that incorporates feedback and ideas from across the ecosystem. Some stakeholders have raised concerns that the taxonomy may not be granular enough; some have suggested the taxonomy should account for regional and country-level variations.

• What taxonomy should be used? Who should create and maintain it?
• What standard might be used to determine sensitive categories? Further discussion on the Topics explainer repo.

# Website classification

Topics are inferred by Chrome, using a classifier model that maps site hostnames to topics. The public can inspect the classifier—either by downloading it locally, using the Topics colab, or utilizing chrome://topics-internals. Some stakeholders have shared individual examples of "miscategorized sites." Others have suggested that categorization at the hostname level does not effectively assign topics for sites with diverse sets of content.

• Should sites be able to provide their own topics? Further discussion here.
• What should happen if a site disagrees with the topics assigned to it by the browser?
• Should the classifier consider additional data beyond hostname (for example, page URL, page content)?

# Topics ranking

The top five topics for an epoch are selected based on frequency. That is, the browser selects the five topics that appeared most frequently in a user's browsing history for a given week. Some stakeholders have shared alternative approaches to calculating the top topics, including variables such as inverse document frequency (also known as TF-IDFA), a notion of commercial value by topics, and the frequency of advertising landing pages on the web.

• What other variables should be considered when choosing the user's top topics? How should those variables be weighted?