From what I've read and been told by someone I know who works in sleep medicine, Oura is very good at detecting when you are asleep and is generally accurate to within 15 minutes of a full polysomnography study. However, I've been told to basically ignore the breakdown of sleep stages because that is muddled with noise. I'm not sure if he considers that to be true of the latest generation ring or if he was operating on data from older models.