Or too many people are slurping up anecdotes from the same watering hole that confirms their opinions. Outside of academic papers, I don't think I've ever seen an example of "measuring" output that couldn't also be explained by stochastic variability.