Computer simulations allow the prediction of hydrocarbon volumes, comp
osition and charge timing in undrilled petroleum prospects. Whereas di
fferent models may give different hydrocarbon charge predictions, it h
as now become evident that a dominant cause of erroneous predictions i
s the poor quality of input data. The main culprit for prediction erro
rs is the uncertainty in the initial hydrogen index (H/C) of the sourc
e rock. A 10% uncertainty in the H/C may lead to 50% error in the pred
icted hydrocarbon volumes, and associated gas-oil ratio. Similarly, un
certainties in the maximum burial temperature and the kinetics of hydr
ocarbon generation may lead to 20-50% error. Despite this, charge mode
lling can have great value for the ranking of prospects in the same ar
ea with comparable geological histories.