I agree with the need for more analysis.

This site seems to take a MadLibs style of 'product that exists' for 'company type that exists'. Surely there are legitimate results from that, but not without a lot more work.