In today's market situation it is becoming increasingly important to minimize the risk of making poor investment decisions. Reservoir models, and associated model predictions are often used a direct input in these decisions. Hence, it is of utmost importance that reservoir model predictions and associated uncertainty estimates can be trusted. In order to achieve this task, a clear requirement is that the reservoir models consistently honour both static and dynamic data, while capturing all modelling uncertainties. In this paper we will illustrate how using ensemble based methods can ensure both fast and consistent data integration and give an improved representation of the model and prediction uncertainties, leading to improved reservoir management decisions. Furthermore, we illustrate that the traditional, base case, approach leads to the wrong investment decision as model uncertainties are ignored.