I read this on an article on HN (I don't recall exactly where), but it was a useful perspective on Agile.

Essentially it argued that when people "analyze" the effectiveness of Agile they often choose a "as-treated" approach instead of an "intent-to-treat" approach. That is wrong. If Agile is supposedly a treatment for for the ills of modern software development, and 95% of the teams trying to do agile are failing, then that isn't a problem with people thats a problem with agile and the agile manifesto. If Agile is so vague that almost noone team can get agreement to do the good parts, and instead they all just do scrum (which is deemed mostly bad) because only Scrum is detailed enough to follow, then again that's a strike against Agile.

This eliminates all of the No-True-Scottsman Agile issues. Agile should be judged by the amount of people trying to do Agile and failing. Otherwise it's not different than "abstinence only" contraception. Yes it works if you adhere to it, but majority of people fail to adhere to it. That's not a failure of people, that's a failure of the prescribed solution. A better solution is condoms.

Either make Agile prescriptive (which defeats the point of Agile), or declare Agile almost valueless. Until someone comes up with a flavor of Agile to counter Scrum, that actually follows the Agile Manifesto it's not worthwhile to discuss Agile.

Someone needs to come up with an agile method that helps teams operate in nightmarishly bureaucratic orgs, wherr 3/4ths of your work is spent just clearing the way to build something of value.