This question was recently posted on Quora. Below is my answer.
It depends what kind of data scientist you want to become. I think many university curricula include material that is advanced but that you don’t really need. Also, they offer not enough practical, professional coding and big data manipulation that would help you right away when starting a career. It also depends on your background: mine was math, stats, data analysis, and applied computer science, so the transition was obvious and just simply about changing my job title.
You can do great implementations and analyses even if you know little statistics. Most of my data science projects (for instance image analysis or automated stock trading) involved designing algorithms, some doing the same thing as complex statistical techniques such as random forests, yet the algorithms themselves could be understood without statistical or mathematical knowledge, and could as well have been designed by software engineers who do not usually call themselves data scientists. It is particularly true if you design tools to beat your competitors: all of them are using the same techniques; if you come up with something original that works, you are more likely to successfully compete with them, whether your tool relies on advanced statistical models, or no statistical model at all. In a different context, I performed non-standard statistical testing in a way that does not involve using probability theory: it was all data-driven (bottom-up approach) rather than theory-driven (top-down.)
I find sales to be more difficult than data science, and more rewarding financially. Of course, it depends on what you are selling, but selling data science software to a large corporation for instance, is more difficult than succeeding as a data scientist working for that same company, in my opinion.