Marketers Can Glean Private Data on Facebook
SAN FRANCISCO — Online advertising offers marketers the chance to aim ads at very specific groups of people — say, golf players in Illinois who make more than $150,000 a year and vacation in Hawaii. — But two recent academic papers show …
Context & Ripple Effects
Facebook's advertising business has been built around precision since its 2007 ad targeting plan, when the company first promised marketers audiences sliced by demographics and declared interests — a capability rival MySpace was racing to match at the same time. Two new academic papers argue that same machinery cuts both ways: because ads can be aimed so narrowly, the act of targeting itself reveals whether a user fits a profile, letting marketers probe for attributes users never volunteered.
The finding lands during a bruising October for Facebook on privacy: a Wall Street Journal investigation days earlier found popular applications transmitting users' personal identification information, and the company had just adjusted its Photo Memories feature after complaints about ex-partners resurfacing in the sidebar.
First-order effects
- Marketers gain a low-cost research channel: precisely targeted ads double as yes/no tests of whether any Facebook user matches a sensitive profile, such as the income-and-vacation segments described in the papers.
Second-order effects
- Facebook faces pressure to cap how granular its ad targeting options can get, since every additional targeting axis is another attribute third parties can infer — a constraint it resisted when rivals like MySpace were competing on the same precision.
Third-order effects
- If ad-platform precision keeps expanding while inference techniques mature, the boundary between 'data users shared with advertisers' and 'data extracted from the ads themselves' erodes, making microtargeting a recurring flashpoint between ad businesses and privacy regulators rather than a one-off controversy.
The trend: Advertising platforms are discovering that targeting precision is itself a data channel, turning every refinement of audience segmentation into a potential privacy exposure.