That's good because either way it costs them money.
Also fake data being more chaotic would be good because it would ultimately cause them to over filter and throw out real chaotic data, effectively over fitting their marketing efforts to slightly more average people.
Additionally, a good arms race would be fun here. Keep making the data more and more realistic each generation. Add an option to use plugin datasets so wallstreetbets can crowdsource fucking with investment banks (ie why is everyone suddenly spending all their time at targets?)
... which is fine. Incentives to refine data analysis are always welcome.