Announcing Novi Causal Modeling

Optimizing your spacing/stacking design, especially with complex parent-child relationships, is critical to deliver returns across cycles. It’s also one of the hardest problems to analyze. That’s why we are excited to announce the release of Novi’s newest algorithm, Causal Modeling, designed specifically to generate powerful results for spacing and infill scenario analysis. 

Drawing on principles from the fields of public health and economics, Novi data scientists have developed Causal Models that give greatly improved spacing sensitivity without sacrificing model accuracy. So how does it work? And why should you care?

The Importance of Spacing

Why is spacing optimization so important? Take as an example a single 2 mile by 1 mile unit in Martin County in the Midland Basin. Here, you could reasonably develop various combinations of the Wolfcamp B, Wolfcamp A, and Lower Spraberry, to say nothing of more exploratory zones like the Jo Mill or Dean. Each of those zones could be developed with anywhere from 4 to 12 wells. All zones could be co-developed, or you might start with the Wolfcamp A and Lower Spraberry before coming back to develop the Wolfcamp B.

Furthermore, you could stagger your wells, stack them, or some combination thereof. When you add in existing parent-child relationships,  plus the range of possible completions designs, it’s easy to see how complex the problem is, with thousands of scenarios to be analyzed. At $6-10 million per well, you’re talking a huge CapEx range from ~$50 million at the low side to $250 million or more at the high side. Drill your wells too tightly together, and that huge amount of capital is wasted. Space your wells too far apart, and it’s easy to leave millions on the table. 

With so many variables in play and massive capital expenditures at stake, it’s hard to overstate the significance–and the inherent risks–of optimizing spacing and stacking in unconventional developments. 

The Challenge of Spacing Analysis

Reservoir engineers who have unconventional experience know just how challenging forecasting spacing and parent-child relationships can be. While machine learning excels at complex, multivariate problems like we have in oil and gas, spacing can be the toughest challenge for these algorithms. This is because operators tend to downspace the tightest in the best rock, which can have the effect of “covering up” the effect of spacing with rock quality or completions intensity.

But not to worry! Recent developments in other fields have provided techniques to successfully model treatments in these difficult data structures. Imagine you are a public policy expert evaluating the impact of introducing healthy grocery stores into neighborhoods without one (known as “food deserts”). Is this a worthwhile project for a government to subsidize, or would tax dollars be spent better on other efforts? If you just looked at the data, you’d see a clear relationship between healthy food stores and healthy outcomes, but many other factors come into play: income levels, school quality, environmental pollution, etc. 

For this class of statistical problems with strongly correlated features, researchers have developed an approach called causal inference, with the goal of accurately estimating treatment effects on a population, even with confounding and interacting variables.  Returning to our food deserts example, causal inference would estimate the impact of introducing a healthy grocery store independent of the other variables.

There are various approaches under the causal inference umbrella, but we use “double machine learning”. Without giving too many of our (patent-pending) details away, this method allows us to disentangle the true contribution of pad design from local geological features, and learn different spacing effects in different acreage, or for different kinds of completion intensities. This enables Causal Models to pick up subtle relationships, make accurate spacing degradation forecasts even where the data are limited, and generally improve spacing sensitivity. 

The Power of Causal Models

For operators or investors in the unconventional space, Causal Models can help improve returns on development designs and improve forecast accuracy for complex developments, even where data becomes sparse. 

One distinguishing feature of Causal Models is their ability to capture the time-variant impacts of spacing on production. Often, dense developments look just as good as widely-spaced developments in early time, only for their production to fall off a cliff. Understanding when spacing makes an impact is critical to your development strategy.

Causal Models excel at forecasting across the intricacies of multi-zone, parent-child developments. Novi customers are using Causal Models today to optimize distances to parent wells, stacking/staggering patterns, and development ordering. For instance: should you co-develop the Wolfcamp B or come back and drill it in two years? This choice will not just affect your Wolfcamp B wells; Wolfcamp A wells are also impacted by developments in surrounding zones, and your forecasting must take this into account. 

For spacing decisions, the value added or lost on a single unit can be in the millions. Using Causal Models, we examined potential development scenarios for a unit in Martin County. Here, I’m plotting total unit NPV against the number of wells developed per section (“WPS”), a common chart used by operators for decision making. The Causal Model results show maximum NPV at 18 wells per section. The optimal development generates $7MM more in NPV compared to the next-less-tight scenario; overtightening to 24 wells per section erodes $9MM in value. 

Traditional methods for this type of analysis are time consuming, error prone, and subject to human bias. Novi Causal Models both improve your engineering team’s efficiency and give you a powerful recommendation for incredibly impactful development decisions. 

Conclusions

For operators and investors navigating the complexities of unconventional developments, Causal Models offer a powerful blend of precision and practicality. They serve as a comprehensive toolkit for intelligent scenario analysis for spacing/stacking and parent-child scenarios. When you combine them with Novi’s best-in-class data offering, you truly have an industry-leading data to decision workflow. 

Live Webinar + Q&A Session

Curious About Novi’s Causal Modeling Algorithm?

In this live webinar, you will learn how Novi’s new algorithm improves model sensitivity for spacing and parent-child scenarios, providing powerful results for previously difficult-to-analyze problems.

Ted Cross, our VP of Product Management, will discuss improvements to our analytics offerings and how the causal modeling algorithm enhances spacing and infill scenario analysis while maintaining model accuracy.

Save your Seat

Ready to see what Novi Labs can do for your team?

You can always request a free demo to see what’s under the hood and how Novi Labs can help you to unlock hidden investment opportunities.

Ted Cross

Ted Cross is VP of Product Management at Novi, where he builds powerful machine learning data and analytics products to solve the toughest problems in unconventional oil and gas. Prior to Novi, Ted was a Senior Geologist with ConocoPhillips.