A Black Box? Machine Learning to Combat Regulatory Uncertainty

Natural Gas Projects
June 16, 2017


The energy supply chain has long recognized the value of collecting and analyzing data to improve operational efficiencies. Rig counts, sensors, and weather forecasts are just a few of the tools managers typically employ to increase transparency and create competitive advantages. One inherent challenge for energy markets, however, is operating within a highly-regulated permitting process which, for capital intensive projects, can include a feedback loop of federal, state, and outside agency interaction. Over the past year, customers of LawIQ’s gas pipeline project predictive modeling have received the output from our models, which combines cutting-edge statistical analysis with industry’s-leading subject matter expertise. With billions of dollars of capital at stake, how did we perform? 


Harnessing Predictive Modeling to Yield Value


For the fourth consecutive year, oil and gas executives marked regulatory scrutiny as the top risk to their organizations. Regulatory uncertainty creates a complex and acute “blind spot” associated with project planning, bringing risk associated with timelines required before pipelines, power generation facilities, and electric transmission lines are able to be placed into service. Such blind spots also create financial uncertainties for company revenue and cash flows, leaving all stakeholders, from engineering and construction contractors and developers to financial backers, exposed. 


With expectations from partners and financial markets, the hedge against these unknown risks is often a buffeted schedule to provide stakeholders maximum flexibility to deal with regulators and unforeseen roadblocks. In fact, data shows that, historically, company expectations for FERC approval dates are on average 90 days early, and deltas (i.e., difference between actual event and company guidance) yield a standard deviation of 125 days, while company in-service date forecasts are on average 58 days early and those deltas yield a standard deviation of 145 days. These variations are informative for projects with upcoming deadlines (see: Special Report: NEXUS sticks to in service date) or the potential impacts on the date molecules will flow (see: A Win-Win?: Early In Service).


Fortunately, FERC-regulated projects are, by virtue of a standardized process, more similar than different. Company filings, regulatory requests, third party opposition (and support), federal and state permits/certificates, and hundreds of other common attributes yield necessary inputs to construct objective, statistically-significant trends and correlations. Machine learning methods can detect subtle patterns in these historical input data points and yield real-time forecasts for projects still undergoing the regulatory process. LawIQ’s predictive model — with outputs next week available via the web-based platform — forecasts the major milestones, including regulator approval and project in-service dates.  


Modeling Results and Performance


Two years of predicted events allow us to evaluate modeling of forecasted events relative to actual outcomes. The results are clear. As shown in the graphic below, which includes projects filed in 2015, predictive models offer objective transparency and dramatically improved certainty to the expected receipt of one such milestone, the FERC certificate decision. Improving the expected timing for this event allows for better planning, both initially when the application is filed, and over time as more inputs are gathered, because the model automatically adjusts to improve certainty. 


Supplementing Industry Expertise 


But, of course, even the most sophisticated models designed by industry experts can have limits, oftentimes bounded by the relevance of qualitative analyses color. As such, our team of professionals offers qualitative analyses, such as those involving, for example, the fine nuances related to the ongoing tensions between pipeline companies, such as Energy Transfer, and state and federal regulators. The vagaries of regulatory permitting can be demystified, with substantial upside and risk mitigation, depending upon your business, for those willing to leverage best-in-class technology with subject matter experts. Analyses must also account for one-of-kind events, such as the FERC’s lack of quorum, which should be resolved soon.  


This qualitative context, when coupled with objective statistical analysis, produces a powerful set of actionable intelligence that is scalable, repeatable, and measurable. As the energy infrastructure build-out continues, LawIQ keeps pace to grow our solutions to meet the needs of our customers.