Many types of risk analysis approaches are used and if the management of a company is successful with one approach it may be hard to justify changing it. One example of an approach that is not recommended, yet still relatively commonplace is the adjustment of discount rates to account for risk. A manager may wish to see infill wells and workovers evaluated at an NPV10, while exploration wells would be evaluated on NPV25. Intermediate levels of risk might be evaluated with more discount rates. In principle, if the reservoir engineers and analysts conducting the evaluations use consistent approaches for estimation, the adjusted discount rates might be correct. I am generally skeptical that this is the case.
If we revisited the “drill vs farmout” case presented earlier, it can be left to the reader to evaluate the cases at 25% and decide when it is preferable to drill or farmout. The “risk the discount rate” approach often incorrectly assesses risk when comparing cases. Rather, it is strongly recommended that the “risk” be either overtly applied to cash flows or the Monte Carlo simulation approaches be used for risk analysis.
A common approach to risk and the handling of unknowns include sensitivity analyses in which relevant input parameters are modified and the impact of these changed assumptions are displayed as a function of the changed parameter(s). Previous figures such as those that showed different cash flows or economic results for different well spacings are typical. In many cases, one parameter may be modified that actually is not independent of the other parameters. For example, in the following tornado chart the initial rate is varied as is the net thickness. There can be many reasons that wells with the same thickness have different initial rates; the varying rates may be a function of permeability, viscosity, skin, etc. This particular type of chart is often used to quickly show the most important factors driving variations (in this case for NPV10) in economic value.
Simply varying one variable may have complex results that are not always handled consistently in such sensitivities. If the engineer introduces a net-to-gross sensitivity, the hydrocarbons in place are obviously changed. What about the initial rates? Interaction of the aquifer? Fluid handling and artificial lift issues? The simple sensitivity introduces numerous assumptions that need to be handled in a consistent and easily understood manner. This is particularly important when conducting sensitivity analyses of reservoir simulation forecasts for reservoirs with significant production histories. Consider a case in which a very good history match is believed to have been obtained. In such a case there might still be questions about many of the parameters. For example, if the engineer simply changed the residual oil saturation as a sensitivity case without reconstructing the history match, unrealistic variations in future recoveries are likely. Had the engineer used a different value for Sor and “re-matched” the cases, it is likely that the forecast results would show less variation from the history matched case than would a forecast that simply changed the value for Sor, exaggerating the sensitivity of the reservoir simulation to errors in assumptions.