I find it to be alright...I don't think its anything truly groundbreaking over traditional MOS. What can be very valuable about it is how it provides temperatures based on percentile which can be a huge asset in highly anomalous patterns. For example, I think it was a couple years ago when they were getting very big heat in the West I remember some places where the 90th percentile which was verifying over the typical median which is spit out.
I think it absolutely sucks though along boundaries, which I mean can be expected to a degree, but it almost always, always to favor towards the warmer side of the boundary. I've seen NBM bust by like 15F because it had the warm front blowing through and that never happened.
I also like how it does not have the limit of reporting cloud cover...so it will pick up on the high clouds while traditional MOS will spit out CLR