Dibbern's example is excellent. He doesn't 'know', of course, but that is not the point. Almost all predictions of earnings will be wrong, but the purpose of a prediction is not to be 'correct', but to support forward-looking decisions.
A prediction is simply an estimate. Initially, they may be no more than guesses, though one can get better at making estimates by reviewing past estimates against actual performance.
And the real value of good estimates is not in deriving any satisfaction of accuracy - there will always be uncontrollable factors causing variations. Rather, the value is in supporting your decision-making as to where to invest your time, ie: ROI.
Ultimately, the benefit of dibbern's approach will be to maximise the revenue being earned from the effort invested.