Abstract

The oil and gas industry, all industries, are moving toward analytics, machine learning and artificial intelligence. Studies show that 70–90% of the transitional efforts toward these technologies fail. Management, data analysts, and domain experts must all work cooperatively, in conjunction, with understood objectives and with understanding of the successful use, failure, and misuse of these tools. If one part of the system fails; the entire system fails. The larger the company, likely the larger the initial efforts (e.g., bigger data), and more likely the lack of successful use and implementation.

Having brought geologic, engineering, and economic analytics, machine leering, and artificial intelligence into several organizations, the lessons learned are to start small with well-defined problems, quick analytical solutions, and quick operational results to demonstrate the superior results of these technologies. Several onshore Gulf of Mexico examples from operating, upstream oil and gas companies are provided from smaller, well-understood issues to building best-of-class, basin-wide, geologic/reservoir characterization, economic and outlook modeling of multiple unconventional resource plays using dozens of variables with billions of input data, big data, machine learning, and artificial intelligence.

A simple model of a traditional Paluxy Sandstone reservoir, Martinville Field, Simpson County, Mississippi (Fig. 1), will be discussed to demonstrate how a well constrained problem, small number of variables, and a well-designed model, can be used to prove the power of analytical models and solve differences of opinion and judgement of domain experts and decision makers with various objectives. Another beauty of the model is that since it was used as an introductory model for a major operator in a case heard before the State Oil & Gas Board of Alabama decades ago, the accuracy of the model verses the domain experts and decision makers is proven by the test, a look back, in time.