From Bitnami MediaWiki
Revision as of 18:45, 5 February 2021 by Jkoran (talk | contribs)
Jump to navigation Jump to search

The goal of Google's Privacy Sandbox is prevent marketers from engaging a particular audience in a particular context.[1]

TURTLEDOVE (“Two Uncorrelated Requests, Then Locally-Executed Decision On Victory”) proposal attempts to enable consumer retargeting without any marketers receiving any identifier.[2]

The proposed mechanism requires the web client to store some the attribute information marketers would want to use as an input into audience creation or auction management. As the name implies, the web client performs two separate local auctions based solely on context or solely on audience information. The winner of these is sent to Fenced Frame ad slot on the publisher.


TURTLEDOVE removes transparency and control from publishers, by requiring the final auction in a given ad slot to be conducted by Google. In addition to removing price transparency and introducing a time-delay to publishers in how well their websites are performing, TURTLEDOVE also impacts the publishers ability to appropriately understanding and managing:

Traffic quality (TQ) - is this a legitimate request from a real user or non-human traffic?

DSP qualification - which DSPs are interested in this request?

Bid solicitation - which DSPs will place a bid on this request?

Ad Quality (AQ) - which bids are eligible given the publisher's requirements?

Valuation - which bid is most valuable to the publisher at this moment?[3]

Another potential impact of removing this information is a degraded end user experience.

Because Turtledove mandates the browser determines the final auction and relies on Fenced Frames, publishers cannot compare in real-time the monetization of the ad slot by Google vs competing ads sourced via other mechanisms. For example, if a publisher has other demand sources, such as an internal sales team, these will need to forward their highest desirable ad ad to browser for it to compare against the winning ad among the context-only ad and the audience-only ad it processes within the auction.

Open Questions

  • What is the impact on web experiences given many client devices do not have the same processing power of the servers that currently select which content to render?
  • How does the local client auction support publisher custom logic, such as factors other than price to select which ad to render?