Software / ABX Engine

Small experiments. Clear decisions.

A look inside consistent visitor assignment for storefront experiments.

Contribution: Product architecture & experiment tooling
Project: Studio project
Explore: allocation / state / experimentation

Try the interaction

Try a visitor ID. Then try it again.

Hands on

ABX Engine — interactive reconstruction preview

Reconstructed interaction · sample data

Interactive reconstruction · sample dataNo account needed. Nothing is sent.

How this works ＋

Derive a stable assignment from a visitor identifier and show the assigned treatment. This small reconstruction demonstrates allocation, not a production analytics service.

The boundary. Stable allocation alone does not establish statistical significance or experiment validity.

Inside the build

The problem worth solving.

An experiment becomes difficult to interpret when a returning visitor sees a different variant unexpectedly.

A considered approach.

Derive a stable assignment from a visitor identifier and show the assigned treatment. This small reconstruction demonstrates allocation, not a production analytics service.

Design decision

Stable allocation alone does not establish statistical significance or experiment validity.

Project in context