Offshore development requires large investments, which have to be made in the presence of multiple sources of uncertainty. Quantification of uncertainty in predictions of a reservoir’s production and, consequently, the project’s revenue alleviates some of the risks and facilitates more informed business decisions. Despite significant advances in the field of uncertainty quantification, it is still common practice in the industry to rely on most likely parameters for the wellbore and pipeline multiphase flow models when making predictions for the project design. We focus on predictive uncertainty of pipe-flow models, which are used to forecast the cumulative production of an oil reservoir whose fluid properties are typically unknown during the exploration phase. The uncertain inputs of a flow model are treated as random variables with a multivariate Gaussian probability density; the model’s prediction of cumulative production is given in term of its distribution, which is estimated via Monte Carlo with Latin hypercube sampling. A global sensitivity analysis is performed to identify the model inputs contributing most to the predictive uncertainty.