But he also has a reputation that has been built up over years through insightful blog posts. Many people in the applied math community know his writings, and there is a general recognition of his expertise. https://www.johndcook.com/blog/
Data science is too general and faddish (I say this as someone whose job title contains that phrase). The market is saturated with too many people who claim they do data science. If someone were to give me their business card and tell me they did general data science work, I probably wouldn't even give them a second look.
My suggestion is to establish yourself in a few problem domains (e.g. data science for agriculture, data science for transportation, etc.) and show tangible results. I think that would improve your marketability.
I grew up around consultants, and as with any kind of consulting business, the advice that was given to me was to work for at least 3 different established companies in your domain of interest for 3 years at a time before setting up an independent consulting shop. Without that, you won't have a sufficient street cred to sustain a stream of business to keep you afloat. (that said, this advice is 15 years old -- the part about 3 years + 3 companies might not be applicable anymore, but I believe the part about building a reputation continues to be true).
Conceptually it isn't that hard.
Putting it in practice is not always easy.
I imagine a data science business would be a services business that you can sell to C-level executives.