Official Store-Locator Registry Official Store-Locator Registry

Overview

Built a monthly branch registry for Mexican retail chains that uses only official sources: it scrapes each brand's own store locator, normalizes locations to a single schema and diffs runs to detect openings and closings.

Highlights

1. 01

   Engineered a declarative adapter framework with six extraction strategies (JSON APIs, embedded JSON, sitemap/JSON-LD, HTML listings, grid sweeps and real-browser XHR capture) and shipped 90 brand adapters that yielded 20,805 accepted branch locations in the September 2026 run.

2. 02

   Enforced a single-source rule: no aggregators or map APIs as location sources, and geocoding only for addresses the brand itself published. The rule is backed by QA gates, frozen fixtures and 206 pytest tests.

3. 03

   Handled anti-bot store locators conservatively, using a real, headed Chrome profile that stops at any CAPTCHA challenge and never solves or retries it. Exported results to Parquet and GeoJSON with a per-batch progress board across 406 registered brands.

Stack

• Python
• httpx
• Playwright
• Pydantic
• Typer
• BeautifulSoup
• JSON-LD
• H3
• Shapely
• Parquet
• GeoJSON
• pytest