Intent to Experiment: Privacy-Preserving Attribution

[Martin Thomson]

Summary:

Attribution is the process of counting outcomes (purchases, etc..) that are linked to advertising.  Attribution is critical to effective use of online advertising as it provides information about how well ads are working, or not.

Today, the advertising industry relies on tracking for attribution, which has very ugly privacy characteristics. Browser support for privacy-preserving attribution (PPA) provides advertisers a way to receive aggregated attribution information about the performance of their ads without tracking. Use of PPA effectively eliminates one key justification advertisers use for tracking people online.

We plan to use our Origin Trial infrastructure for this experiment.  It will only be available for certain sites. Note that because Mozilla is paying the DAP aggregation service costs, so participation is by invitation only.

Bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1900929

Explainer: https://github.com/mozilla/explainers/tree/main/ppa-experiment

Standards Body: W3C (PATCG, soon PATWG)

Platform coverage: All

Preference: dom.private-attribution.submission.enabled

DevTools bug: None currently

Link to standards-positions discussion: https://github.com/mozilla/standards-positions/issues/753

Note that I just closed this with a "defer" position.  This overall effort still has a lot of uncertainty about a number of aspects.  Part of the point of this experiment is to inform our choices about some of these open issues.  Though we're generally positive on the general direction of this work, we generally refuse to provide a position when work from other browsers is similarly ill-formed.

web-platform-tests:

As an experiment that is deliberately proprietary, WPT is not appropriate, even if WPT will be essential for any final API.

Other browsers:

This is our experiment only, but there is related work on aggregated attribution.

Blink: Chrome ships similar functionality with their Attribution Reporting API (the Summary Reports, not Event-Level Reporting), but with effectively no privacy protections. See https://github.com/WICG/attribution-reporting-api
WebKit: This experiment is very close to Apple's design in their private ad measurement API.  See https://github.com/patcg-individual-drafts/private-ad-measurement