OAG Analytics Combines Domain Expert Algorithms and Machine Learning for Optimum Well Spacing
Optimized unconventional well spacing and delighted customers
OAG Analytics, a Texas-based software company that accelerates collaboration between domain experts and data scientists, solved the problem of modeling production for oil and gas wells and basins. OAG's solutions uniquely combine reservoir physics, machine learning (ML), and data visualization software to deliver transparent and interpretable hybrid ML models. This has automated the well spacing process for multiple operators, five times faster and at one 10th the cost of an internal effort.
Addressing well spacing
Determining well designs and spacing between wells is both extremely complex and a huge opportunity. Until OAG's hybrid ML innovation, neither physics alone nor ML alone materially improved pre-drill production predictability. Across a drilling unit, the goal is to balance frac intensity and well spacing to maximize barrels produced per dollar spent. "If you put the wells too close to each other, it's possible to destroy $100 million of value in a single day," says OAG Analytics Founder and CEO Luther Birdzell. "That's why Wall Street is concerned about it. That's why the industry is concerned about it. And that's why we've had a laser focus on this problem since 2013."
Mr. Birdzell observed that Wall Street, particularly for shale oil producers, clearly favors companies that excel through the use of technology. "In August 2019, one company that came in just a skosh under its production forecast lost almost 30% of its stock value in a single day. It's not a coincidence that the company had a well spacing test that didn’t go as well as they had forecast."
OAG Analytics was founded with the vision to address the hardest problems in the industry. Its technology includes a proprietary reservoir and geoscience algorithms, AutoML, unlimited storage and compute on AWS, and integrations with TIBCO Spotfire software. "TIBCO was the first partnership we built in late 2013, and we've been with TIBCO every step of the way," said Birdzell.
Creating high-value estimates
OAG's approach starts with automation to liberate data from silos and make it 12 times faster to build and perform quality control on a dataset from a large data library. Its software provides data science workflows for training and validating models as well as purpose-built applications for domain experts to easily interact with OAG's hybrid models. OAG insights are also available through its integration with Spotfire analytics.
OAG workflows are complementary to industry-standard techniques for estimating production. A type curve is typically the average production of manually identified similar wells in a 25–100 square mile area. Machine learning algorithms tend to converge around a mean or median, resulting in faster but not better type curves. "Higher fidelity pre-drill insights are uniquely possible by combining science with machine learning," says Birdzell.
Serving market demand
Initially, OAG Analytics believed that geoscientists and petroleum engineers would want to do their own machine learning, data engineering, and analytics (the citizen data science model). Instead, the industry is adopting solutions that accelerate collaboration between data scientists writing Python code and subject matter experts consuming those models through data visualizations. The company pivoted to an extensible SaaS model in 2019, which includes a well spacing solution that can be deployed in less than eight weeks—and has done so for every major unconventional basin in the US, over three million acres.
OAG also partners with its customers to help them develop and deploy their own IP on the OAG platform. This accelerates asset teams' ability to better respond to rapid changes in the market.
"A couple of years ago, it went from 'How does this work?' to 'How quickly can you make this work for me?'," says Birdzell. "We have identified billions of dollars of annual cost savings for our oil and gas customers."
Developing insight and delight
"Extensible SaaS business models enabled by technology partners like TIBCO and Amazon are fundamentally changing the enterprise software experience from something that is tolerable—some days maybe not even tolerable—to something that is truly delightful."