This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 179996, “Unconventional Risk and Uncertainty: Show Me What Success Looks Like,” by David S. Fulford, Apache, prepared for the 2016 SPE/IAEE Hydrocarbon Economics and Evaluation Symposium, Houston, 17–18 May. The paper has not been peer reviewed.
This paper presents approaches for proper risking of uncertain recoverable volumes for an unconventional resource, taking into account the chance of false positives from appraisal-well information. A subjective risk tolerance can be included to respect how aggressive or conservative a company may be in pursuit of a project. A method of recasting production profiles is demonstrated as an improvement over the common method of scaling, or factoring, the initial production rate.
Introduction
Onshore unconventional liquids-shale and gas-shale plays are often referred to as “continuous accumulations” or “resource plays.” The characterization of perceived geologic continuity frames perception of these plays as having little to no risk, to the point that “well manufacturing” is a noncontroversial term used to refer to development of these assets. Exploration-and-production companies often tout a “type well” as an indication of the viability of their assets. However, a type well, by definition, represents the mean of a population of wells. Not all individual wells must be commercial in order for the mean well to be commercial.
The well population is something that cannot be predicted, but only estimated. This implies uncertainty. However, the argument that there is risk remains to be made. Regardless of the cause of this no-risk assumption, the tools exist to address uncertainty. Empirical modeling must follow from first principles. This applies not only to the evaluation of individual wells to determine their ultimate recovery but also to the statistical models built to judge the economic value of drilled wells.
In the analysis of small sample sizes (and assuming independent and identically distributed samples), the variance is always overestimated. The cause of this phenomenon lies within the roots of regression analysis—or, more accurately, in the assumptions made during the application of regression analysis.
