When we first started flying FMC & PMS equipped aircraft it was SOP to cross check the wunderkind box numbers against dispatch data and QRH human generated results.
A recommendation I read long ago suggested two-pass entry for safety critical human-typed data into the management computers. The paper also included the idea of two-crew entry measures with separate source data streams. Each method could have caught some but not all slip of the finger/byte errors. The final suggestion was thrown out due to the numerous minor-difference tail chasing exercises the process would have required to stay legal.
Waste of time for the statistical results was the final verdict. The possibility of bad numbers shared upstream fouling the entire system was one of the key arguments against the proposed two-stream procedures.