I used to work for a relatively large company. (Heavy machinery that is yellow and named after an animal). Some of their projects were many-many-millions of dollars in value and investment. I was the only PhD statistician in the entire division. The rest of their "data scientists" were re-purposed engineers. Yes, they knew how to make Python or R give them results; no, they had no clue how to actually explain why the test they used was correct.

One fast example. We were working on a project that would alter the manufacturing process of about 40% of the parts this company made. When you have $54 BILLION of revenue per fiscal year, that is a heck of a lot of parts. When I joined the team, they were only a few months away from implementing a initiative to change the manufacturing processes. They had come to the conclusion that this was correct by running a 2-sample t-test. (So the basic undergrad test). The only issue was that all of these self taught statisticians were completely oblivious to the fact that the t-test assumes the samples come from a normally distributed population. They also overlooked the fact that the two population groups being compared did not even have close to the same variance values. And they were about to make a several million dollar decision off of this.

Turns out that a non-parametric Mann-Wald-Whitney test completely destroyed everything they had done.

But these employees were self trained, so that's a win for the company! The company avoided having to pay someone $40k more per year to save that measly $5 Million. (And the costs could have been MUCH more if it led to loss of customers, propagated engineering failures, faulty equipment, etc.)

This is the definition of pennywise and pound foolish.