Diff-in-Diff: Privacy-First Causal Inference

How difference-in-differences can estimate causal impact without tracking individuals—essential in today’s privacy-first world.

One underrated benefit of difference-in-differences methods used to estimate causal impact is that they don’t require tracking the same individuals over time - a key advantage in our shift toward a privacy-first world.

Diff-in-diff works by comparing average outcomes between two groups over time. Nothing in the formula requires tracking the same individuals. Different random samples from the same group across time periods still work.

This makes diff-in-diff - and other aggregate data methods like synthetic controls - perfect for marketing measurement, ad tech, and other fields where privacy concerns are growing.

As we gradually move to aggregate, privacy-first data systems, causal inference tools will be essential for data scientists to stay ahead.