@reiver

The Failure Rate of Scientific Breakthrough Research

If success is defined as a published paper, well then, the success rate of research could easily match or exceed 20% (an 80% failure rate). A great deal of modern research in fact consists of measuring something to slightly greater accuracy ("measuring X to another decimal point" is the standard put down) or calculating some theoretical quantity to slightly greater accuracy or detail. This sort of research may well have a 20% success rate (or higher, which the author suspects). Genuine breakthrough research, however, may well have a failure rate exceeding 99% or indeed, more accurately, have a 100% failure rate until a key enabling technology or method is developed, something obvious only in retrospect.

-- John F. McGowan

from "The Manhattan Project Considered as a Fluke"

Quoted on Wed Jun 22nd, 2011