How does Facebook define “sensitive” boundaries for ad targeting?

This blog is full of posts about the boundaries of ad targeting — what kind of use profile data is and should be considered “off limits” for online marketing. Across major ad companies and industry organizations you’ll find important differences in how the lines are drawn. I see this as a defining issue for the success or failure of self-regulation. Doing right by web users will require ad companies to forego significant revenue, particularly when it comes to health and financial ads.

I want to add another definition of boundaries to the mix: Facebook’s ad guidelines, which say:

Ad text may not assert or imply, directly or indirectly, within the ad content or by targeting, a user’s personal characteristics within the following categories:

- race or ethnic origin;
- religion or philosophical belief;
- age;
- sexual orientation or sexual life;
- gender identity;
- disability or medical condition (including physical or mental health);
- financial status or information;
- membership in a trade union; and
- criminal record.

These categories are familiar, and the list is commendably comprehensive (far more so than the  standards from industry groups like the Network Advertising Initiative). What matters most is that the targeting itself can’t “assert or imply” these characteristics, “directly or indirectly.” It reminds me of how Google defines boundaries when they say, for ad networks certified for Google’s exchange, that neither ”inferred or actual” medical conditions may be used for targeting.

Now, it’s an entirely separate question of how Facebook or Google actually enforce these standards on other advertising companies that they allow to reach users in their ecosystems. More on that soon.