TL;DR — Off-the-shelf Amazon research tools hide their decision logic behind a single "opportunity score." This post walks through building your own six-metric niche research pipeline in Python, using ASIN-level data from Pangolinfo Scrape API. Code is production-ready and runs async for cost efficiency. Why this exists I've worked on data tooling for several Amazon-native brands. The recurring frustration was always the same — the team would buy a SaaS tool, get an opportunity score for a category, and have no way to interrogate why the score was what it was. When the inevitable wrong sourcing call happened, there was no decision log to learn from. The fix is to drop one layer below the SaaS tools, pull ASIN-level data directly, and build your own decision framework in code. Amazon niche research data analysis done at this layer is reproducible, transparent, and version-controllable. Once the pipeline exists, every subsequent category analysis costs near-zero engineering time.…