Using AI to understand a market in 60 minutes

I wanted to understand the secondary market for a specific consumer product on Amazon Japan—who’s selling it, which models are available, and how pricing varies by variant.

The old way: manually click dozens of listings, copy-paste into a spreadsheet, then spend hours cleaning and structuring the data. 

What I did instead: asked Claude Code to generate a scraper that collected search results across five queries (first 3 pages each) and normalized them into a structured dataset.

In about an hour, I went from question → dataset, including:

• Hundreds of listings captured across the query set
• Model/variant breakdown and price distribution
• A seller landscape sample (limited by Amazon blocking)
• A clean CSV file ready for analysis

The interesting part: no manual coding and no expensive model. Claude Code wrote the script, I ran it, then I sanity-checked the results and had Claude tweak the script where necessary (e.g., subcategories of products that were misidentified). Most of the time I spent was on taking the output and molding it into a cohesive, audience-appropriate story. 

This is where I’ve found AI tools to be genuinely useful: not replacing analysis or judgment, but removing the tedious data collection step so I can get to the insights faster. 

The code is reusable too. I can run this weekly to track how the market shifts.

If you’re spending hours on repetitive data collection, there’s probably a faster way.

(Originally published on LinkedIn, here.)