There is some axiom in statistics, I don't remember what it's called, something along the lines of "the probability of an event changes the more times it occurs." That's not stating it well, but I can't do any better at the moment. Basically meaning that the stats say that going for 4th and 2 works x% of the time, however the more teams go for it on 4th and 2, that percentage can very well change. In other words, the probability of success has been dependent on the sample size. Other variables would factor in as well, such as how often defenses defend such a thing. The more they see it, perhaps the better they get at stopping it.
I wish I knew how "the analytics" were calculated. There are plenty smarter people than me out there and the NFL surely has the money to pay them. Otoh, some of the decisions make me wonder if they're made in a vacuum.