The two main statistical languages today are S (in the form of R) and SAS, which today pretty much have the form they had 25 years ago.

Whatever usability problems or worker productivity problems they had then, they still have today.

I'm a data language designer, and I look at, largely, four aspects:

1. Usability (learning curve & readability - here Python scores high)
2. Productivity (how long it takes to finish your work)
3. Flexibility (SAS and R don't have problems here, but a macro library will)
4. Reliability (in the QA/reproducibility sense, usually a PL does better than a GUI here)

By the way, I have a language that can produce complex statistical tables much faster than SAS (like 25 lines of code instead of several hundred lines of code). And I'm going to produce a language for data cleaning that will be great for usability (it'll be my third).