Yeah, ideally I suppose we'd have both and perhaps other sorts of measurements, the server logs can still have error because of eg bots scraping porn sites, the uncertainty of attribution, and people who consume porn offline (by downloading it, by ordering DVDs or magazines, etc).

There's also the matter that, from what I've observed, women often prefer erotic literature. That seems very difficult to parse out from traffic.

For what it's worth, I did come across one article where researchers were using data from an analytics company to probe porn use[1], but it wasn't measuring what I was looking for, so I didn't look too closely.

[1] https://www.tandfonline.com/doi/full/10.1080/00224499.2023.2...