AI Competitor Product Tracker: 400+ Pages to One Spec Sheet

In the client's words: "I've only got mine to be 90% accurate and having to manually double check if the information is right takes too much time."

A home appliance brand in Singapore tracks 120 competitor vacuums: 400+ pages on Shopee, Lazada and 10+ brand websites, with about 40 specs each. Linking AI to the sheet, then feeding it PDFs, still left errors. Shopee and Lazada block most bots, and some specs only appear in product photos.

One AI system with a simple dashboard:

Paste the product links and click Start

Many pages open at once, pausing if a CAPTCHA appears

AI reads the text and the product photos

Check the results in a table and export to Excel

The plan started with 2 websites, then grew to 10+ brand sites, each one different, so I built one AI system that can read any page. Then the sources disagreed: Shopee said 120W for one vacuum, Lazada said 150W. The client's rule is now built in: brand site first, then Shopee, then Lazada, lowest price wins, and a spec it can't find stays blank.

2 to 3 minutes per product, with several running at once

41 specs per product, including specs read from photos

1 to 2 cents of AI cost per product

5-star review: "He went above and beyond"

Tech: Python, Gemini AI, nodriver, Streamlit

Full case study: https://mranas.dev/work/competitor-product-tracker