[logan__dev]

I am the author, and I am 100% with you. This data is from scrapped job offers, as said in the article, not from devs posting their current salary. Also some offers have lack of transparency as you said, and some also have malformed salary ranges (which I try to clean). So yeah, this is the exact result of those scraped job offers, but take it with a grain of salt.

[tomhallett]

Totally makes sense. Is there any way for your scanner to filter on "job req from direct company", which would filter out reqs from contingency recruiters?

[logan__dev]

Hmm could try something, Glassdoor, Indeed and Dice have a lot of them, so I could filter out them or put less weigth on their job offers. What also needs to be solved, is, what is considered to be a 'Python' job for example. If python is a on a job 'tag', I count it, but many offers have many tags with many different languages and frameworks, and that is no good for the study. Maybe in future, I would just count what languages/stack is specified on the title. Just saying. Thanks for your comments, I would try to improve the next study I make.