Many times we are faced with the business decision of whether or not to develop a sand that is at the limit of seismic resolution and near the noise level of the data. The critical issue is developing a reasonable certainty that there is enough volume of hydrocarbons to develop. A popular approach is to use Bayesian methods to determine the probability of an economic volume of hydrocarbons being present. A problem with this approach when it is applied to these marginal cases is a bias to the answer. Often, this comes from a relatively strong sophomoric prior constraint on the gross thickness and net-to-gross (N/G) of the sands, imposed to keep the inversion focused on the correct seismic reflector. The data are whispering what the answer should be through the Bayesian apparatus, but this whisper is overwhelmed by the sophomoric prior constraints. We found a simple solution to this problem—run the seismic inversion several times using the output mean of the previous inversion as the input mean of the next inversion. This methodology made the difference, in conjunction with a bandwidth improvement in the seismic data, in proving that a well should be drilled. Unfortunately, the well did encounter an acoustically soft lithology of the predicted gross thickness, but it was a shale—the most likely failure mode as predicted predrill.