Increasing K-Anon window size

[radhika-mani]

Request: We would like to propose increasing the K-Anon window size from 7 to 14 days.

Background & Rationale

Chrome recently announced that K-Anon for rendering creatives will be 50 users over 7 days. Google Ads ran simulations using this threshold to measure the impact on advertiser reach and performance and saw that it would have a severe, negative impact especially on creatives used for dynamic remarketing. This impact is also likely to be more acute with small-to-medium sized businesses that may have more difficulty reaching the k-anon threshold.

For these reasons, we would like to propose increasing the K-Anon window size from 7 to 14 days. We believe this change will improve performance for advertisers while also striking the right balance between privacy and utility.

[tylerdev0]

I agree that increasing the window size would reduce the negative impact on specific creatives, especially in the context of #311 where specific creative sizes would be subject to K-Anon.

[leobierent12]

Comment from CRITEO recommendation team

Rationale :

In our continuous quest to enhance the performance of our platform while safeguarding user privacy, we have undertaken a comprehensive analysis of our current K-anonymity implementation. Our findings indicate that increasing the time window for K-anonymity can offer substantial benefits. This proposal outlines the advantages of extending the time window for K-anonymity, focusing on improving product displays and ensuring user privacy.

Current Challenges:

Currently, our platform operates with a 7-day time window for K-anonymity, which has successfully upheld user privacy. However, it presents certain challenges, particularly for products from partners with limited budgets. As per our analysis:

1. Better Chance for Product Display: Extending the time window for K-anonymity would increase the likelihood of a single product being displayed over an extended period.

2. Enhanced Partner Performance: Products from partners with low budgets are struggling to meet the K-anonymity filter within the existing 7-day timeframe.

Proposed Solution: Extending the Time Window for K-Anonymity:

We propose extending the time window for K-anonymity from 7 days to a more accommodating duration of 14 days. This change will address the following issues:

1. Improved Partner Performance: Extending the time window provides products from partners with limited budgets a better chance to pass the K-anonymity filter. This will significantly enhance partner performance and their overall experience on our platform.

2. Reduced Risk of Click Fallback: Our analysis reveals that by extending the time window, we could potentially save up to 5% of clicks currently at risk with our current implementation. This translates to a more accurate product display, leading to a superior user experience.

3. Preservation of User Privacy: Crucially, this proposal does not involve changing the K-anonymity constraint itself. Therefore, user privacy remains uncompromised. We can ensure that the privacy of our users is guaranteed while enhancing the platform's performance.

Extending the time window for K-anonymity from 7 days to a more flexible duration is a strategic decision that addresses the challenges faced by our partner products and users. This change will improve product visibility, partner performance, and the overall user experience while maintaining our commitment to user privacy. We believe that this adjustment will create a better balance between performance and privacy, ultimately enhancing our platform's effectiveness.

[itaysharfi]

We appreciate your feedback and plan to increase the final K value to 50 over a window of 30 days. We are currently updating our explainers to reflect this change.

[JensenPaul]

Closing as this was published: https://developer.chrome.com/docs/privacy-sandbox/protected-audience-api/feature-status/#k-anonymity