What do the statistics 87%, ~63%, and 99.98% reveal about how easy re-identification really is?

Tiny numbers of attributes uniquely identify almost everyone — the "illusion of anonymity" collapses in practice.

Three benchmark figures:

• 87% of U.S. residents are uniquely identified by just three attributes: birthdate, gender, and ZIP code (the Sweeney result).
• ~60–63% are identifiable from only two: gender and a detailed (full day-month-year) birthdate (the deck cites the figure as roughly 60–63%). Adding ZIP code is what pushes this toward the 87% figure above.
• 99.98% can be re-identified from 15 demographic attributes (the 2019 study by de Montjoye et al., Nature Communications) — even in a heavily incomplete/sampled dataset.

The lesson: re-identification needs far fewer data points than people assume. The "illusion of anonymity" collapses once you count how uniquely a handful of common attributes describe a person — which is why "we only kept a few demographic fields" is not a defense.

Tip: The scary one is 99.98% from 15 fields even when the attacker only has a sample — sampling and incompleteness do not save you.

Go deeper:

• Data re-identification (Wikipedia) — sources the 87% Sweeney result and the de Montjoye studies.