I believe that you will need to make a quantum leap in usability. As pedalpete mentioned, very few people have the skills to process meaningful datasets, even those working in industries where we might expect such skills to be commonplace. The people who do have the mathematical knowledge know how to use the existing software tools.
If this were my idea, I would be looking at smaller niches where I could offer a boneheadedly simple way of doing tasks that are commonplace in a particular industry. I'd focus on accepting the dirtiest, nastiest inputs - excel spreadsheets, tables copied from word documents, tabulated lists - and giving the user a big set of simple preset operations, organised by industry and job role. I'd look further down the food chain to people who might deal with data as part of their job, but have no mathematical background whatsoever.
This might be an odd analogy, but I'm reminded of software synthesisers. Nowadays it's relatively rare that musicians have the technical understanding of synthesis to create their own sounds, even with the most basic tools. Those that do understand synthesis have usually invested a lot of time in learning the interfaces of particular synthesisers and are loathe to discard that investment. Developers are focussing on the unskilled mass market with products that make it easy to access, tweak and combine a big library of preset sounds.