Hacker News Reader

In your opinion, which metrics could be better than # downloads to evaluate PyPI packages at scale?

# of downloads from unique IP would be better for sure.

Also if you have a local cache and just check the version, that won't count.