The Poles of Brand Inaccessibility

This post details the technical process of building an interactive 'Poles of Brand Inaccessibility' explorer. It covers data acquisition from Overture Maps, initial data cleaning and filtering using DuckDB and Parquet files, and the challenges of calculating great-circle distances for a large number of points. The post highlights performance optimizations, including saving intermediate query results and the conceptual approach to making the US 'significantly smaller' for calculation efficiency. It also touches on using ChatGPT for category extraction and the use of Hex notebooks for prototyping and interactive exploration.