Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Alternative way to implement negative targeting #1096

Closed
eysegal opened this issue Mar 24, 2024 · 1 comment
Closed

Alternative way to implement negative targeting #1096

eysegal opened this issue Mar 24, 2024 · 1 comment

Comments

@eysegal
Copy link

eysegal commented Mar 24, 2024

The way that negative interest groups are implemented today is through the additional bids mechanism.
We see several problems with that for Native advertising around the order of ads that are returned.
For example, say our placement displays 3 ads:
ad1, ad2, ad3

They are displayed in this order since ecpm(ad1)>ecpm(ad2)>ecpm(ad3)

Now let's say that ad1 has a suppression rule (negative targeting). Now in the current flow, our ad server will filter ad1 and bring ad4 so the new order is ad2, ad3, ad4. Let's say that this is the optimal order.
If we want to use Protected Audience for suppression, will have to return an additional bid of ad1 with the negative targeting and a fallback item ad4 for that auction. So the new order is now ad4, ad2, ad3 which is not the optimal order.

We were thinking that to solve this, there might be a way to pass the negative interest groups to the ad server, in the contextual ad request, without compromising the users' privacy by using a Bloom-filter (or something similar) that represents the Negative Interest groups for the ad server to query. This can even be very noisy and contain data that are not real negative interest groups, it doesn't matter, as long as the negative interest groups are there. It's even ok that in some low probability it would remove some of the negative interest groups to further enhance privacy.

@michaelkleber
Copy link
Collaborator

Unfortunately, all versions of this idea that we have investigated turn out to enable the server to uniquely identify the user.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants