Edge personalization without writing another Jamstack blog post

Personalization at the CDN edge used to be exotic. It is now a few lines in a middleware file. The interesting question is no longer "can we" but "what happens to the cache."

The mental model: your edge function reads a cookie or a header, decides what variant to render, and rewrites to a path that includes the variant slug. The rewritten path is what the cache keys on. Same content, same URL, but cached per variant. No flush needed when content changes; the variants are addressed independently.

We use this for geographic personalization (currency, language, shipping copy), for soft A/B testing without an external SDK, and for logged-in vs anonymous home variants. The variant set has to be small, three to five, or you blow up your cache cardinality. But within that constraint, edge personalization is now a Tuesday-afternoon feature, not a quarter.