Exploration and production use information that is inevitably mingled with noise and non-uniqueness. One remedy is to use Bayesian inference to build a posterior set of nearly optimal Earth models. Also, decision making can be supported by estimating the value of perfect or imperfect information (VoPI or VoII) about Earth properties. For example, the value of original oil-in-place (OOIP) depends on its volume and sale price, and on facility etc. costs required to bring it to market. Here we introduce a methodology to compute VoPI and VoII that estimates their input distributions by applying successively better geoscientific knowledge and effort. It also generalizes decision-tree analysis to contours and curves that graphically reveal the interaction of decisions and unknown externalities. In particular, from seismic tomography we obtain the probability distributions of Earth properties and their derivatives such as gross rock volume (GRV) and OOIP. Given simple assumptions about the effect of cost, OOIP and externalities on value (so that imperfect value is an acceptable proxy for actual value), we find that VoII is qualitatively similar to VoPI and implies the same decision curves, and both VoPI and VoII are less externality-sensitive when more geoscientific knowledge is applied.