Customer data can support useful insights while exposing people if records are too specific. K-anonymity is a de-identification property: on selected identifying attributes, each record must be indistinguishable from at least k − 1 other records. It is one way to make data less identifying without abandoning analysis. section 2.2.2
Use a broader location
Instead of retaining specific addresses, group users by a broader location category. Records that share that category are less distinctive on location, provided enough records also match on the other selected attributes. section 2.2.2
Apply k-anonymity as part of a wider privacy-first sourcing approach, not as a substitute for it. Collect only data needed for a defined purpose, obtain meaningful consent, and protect data through its lifecycle. De-identification helps reduce the chance of re-identification when datasets are combined; governance and security remain necessary. section 2.3.1 section 2.2.2
