Thousands of wells are drilled and stimulated every year, and much data is generated in the process. How can we use this wealth of information to develop our oil and gas fields most effectively? And will analytics and machine learning preclude the need for traditional scientific inquiry and modeling? A fundamental conundrum is how to understand the relative contribution of the subsurface versus the drilling and completion practices to well performance. Data analytics can help, but not by itself; human thinking is still required.
Two case studies are presented: the Montney play in Canada with 450 wells, and the Marcellus play in the US with 230 wells. In both plays, statistical analysis of large datasets did not suffice, forcing the creation of a novel technique called ‘Outlier Analysis’, in which specific hypotheses are systematically scrutinized against the production data. Emphasis is placed on understanding the very best and very worst wells within the range. The outcome was a step-change improvement in well performance in both plays, which provided confidence for a major investment decision for Canada.
This methodology is rooted in the teachings of some of the great philosophers in history: Locke, Popper, and Kant, reminding us that scientific enquiry still has a role to play. With the evolution of data analytics and digitalization, the economic winners will be those that supplement it with human intelligence, based on sound physical principles and logical thinking.