This investigation presents an approach towards a better understanding of achievable accuracy of fuel consumption predictions of ships and provides an example of how a thorough uncertainty analysis of prediction models can be performed. A generic ship energy systems model is used for the fuel consumption prediction of two reference ships: a RoRo ship and a tanker. The study presents how uncertainties can be categorised and handled in four different phases of a ship's life – from early design to ship operation. Monte Carlo simulations are carried out for two environmental conditions to calculate the mean and uncertainty of the fuel consumption. The results show that the uncertainty in the fuel consumption prediction in a very early phase of the design process is approximately 12%, whereas at a very late phase, it reduces to less than 4%. Finally, the simulation model is applied to a real ship during operation conditions.