Part 4/12:
Testing across platforms like Reddit, IMDb, Wikipedia, Amazon, and the near-trillion-entry Common Crawl showed impressive results. In most cases, MAD 2R outperformed previous methods, revealing millions of unique items while maintaining near-total user privacy—covering 99.9% of users and 97% of item pairs with only a tiny disclose slice. This breakthrough paves the way for smarter, safer online trend analysis that respects user confidentiality.