Host: Society of Petroleum Engineers, Denver Section

Event: April 2026 Technical Happy Hour

Time: April 23, 4-6pm (HH begins at 4, Talk at 430, more Social until 6)

Location: Liberty Energy, 950 17th St, Suite 2400, Denver, CO 80202

Sponsor: Novi Labs

Study Group Category: Data Analytics

Speaker: Brett Sinclair, Research Director, Novi Labs

Title: Meeting the Moment: Using Machine Learning to Assess How the Marcellus and Haynesville will
Supply Rising Gas Demand

Abstract:
One of the key challenges we face when trying to analyze scenarios that rely on forecasting the
production from future wells, is how granular we get in forecasting those wells. Traditionally,
operators or analysts looking at very specific asset packages will have highly fine-tuned Type
Curve Areas to hand-select analogous wells, creating type curves by bench, by area, by
parent/child designation, and many more cuts of the data. When doing higher-level analysis –
how much will a basin produce over time – analysts might use high-level averages of production
per well or per rig (or frac crew) across an entire region or basin.

At Novi, we train machine learning models across dozens of features that include subsurface
geotechnical attributes, wellbore and completion design, well spacing, and how much prior
depletion a future well would experience based on its drainage radius and the amount of
resource extracted thus far.

This provides individual well-level production forecasts custom to each potential location that
can be used for extremely focused analysis or high-level macro analysis and everything in
between. It also gets rid of the machine learning “black box” and explains the variation from
average in a well’s predicted performance due to different geological, spacing, completion, or
depletion features.

In this presentation, we will utilize Novi’s proprietary machine learning-derived inventory
forecasts to analyze how the Marcellus and the Haynesville are each positioned to supply the
increased natural gas demand coming from power generation (largely from data centers) and
LNG export terminals. How much inventory is left, and what is the quality of that inventory?

Which companies are best positioned, not only in the near term but for the foreseeable future?
How have activity levels responded to structural shifts in demand?

We will attempt to show that we don’t have to sacrifice well-level forecast accuracy for scale,
and that a machine learning approach also improves individual forecast accuracy while at the
same time saving time, providing the analyst more scope to analyze the forecast’s impact
instead of in its derivation.

Bio:
Brett Sinclair is a Research Director at Novi Labs, where he focuses on applying the outputs of
Novi’s data and machine learning analytics platform to US upstream oil and gas research.

Prior to Novi, Brett spent 10 years at Kimmeridge Energy in the Investment and Operations
teams. At Kimmeridge, Brett and his team conducted analysis and research from technical,
operational, and investment perspectives, to understand the energy landscape, to develop and
explore investment theses, to underwrite deals, and to directly manage portfolio assets.

Prior to Kimmeridge, Brett worked as a Directional Driller for Baker Hughes, drilling horizontals
in the US Lower 48, the Alaskan North Slope, and Saudi Arabia.

Brett has a bachelor’s degree in mechanical engineering and an MBA, both from the University
of Oklahoma.

Speaker Photo: