_ As unconventional developments mature, the easy decisions disappear first. Early in a play’s life, operators can drill wells in largely undisturbed rock and expect consistent results. Over time, however, new wells are increasingly drilled near existing producers. In the Midland Basin in Texas, this has become the norm rather than the exception. These new wells, commonly referred to as child wells, often underperform expectations. Even when landed in the same formation and completed with similar designs to those of the initial wells, referred to as parent wells, child wells frequently produce less than early-generation type curves would suggest. This underperformance introduces uncertainty into well-spacing decisions, remaining inventory estimates, and capital allocation. Operators know parent–child well interference exists. A harder question to answer is how much it matters at a specific location and whether the impact is large enough to change development plans. This case study presents a practical workflow designed to answer that question quickly and consistently using public data, machine learning, and an empirically calibrated depletion model. Understanding Why Child Wells Underperform Before building a predictive model, it is important to understand the physical processes that cause child wells to degrade. Three mechanisms are typically discussed. - Reservoir-pressure depletion occurs when parent wells lower pressure in the surrounding rock. If a child well is drilled close enough to a parent, part of its drainage volume begins at a reduced pressure, leading to lower productivity. This effect should weaken with distance and is most intuitive where drainage volumes overlap. - Fracture-geometry distortion is often less intuitive but equally important. Hydraulic fractures from a child well do not grow symmetrically when they encounter depleted rock. Instead, fractures tend to grow preferentially toward lower-pressure regions around parent wells. This can reduce effective fracture surface area and direct stimulation into rock that has already been drained. Unlike simple pressure depletion, this mechanism can affect child wells even when traditional drainage volumes do not overlap. - Interwell flow, in which fluids move directly from child wells to parent wells through connected fractures, is theoretically possible but appears to be a minor effect at the basin scale in the Midland. Because it is difficult to observe consistently in public data, it is not explicitly modeled here. The workflow described in this study focuses on the first two mechanisms, which together explain most of the degradation patterns observed across the basin. Establishing an Undepleted Performance Baseline Quantifying depletion requires a clear definition of what a child well would have produced in the absence of nearby parents. Rather than relying on manually selected analog wells, this study uses a neural network trained on all horizontal wells in the Midland region to estimate first-year cumulative oil production. The model incorporates lateral length, stimulation fluid and proppant intensity, well spacing, landing zone, mapped reservoir properties, well orientation, and geographic location. Horizontal depletion is included as an input, allowing the model to learn how proximity to existing wells affects performance. To estimate undepleted performance, each child well is reevaluated with horizontal depletion set to zero while all other inputs remain unchanged. This produces a well-specific baseline representing expected performance in undisturbed rock. Child-well degradation is then defined as the percentage difference between actual first-year oil production and this baseline. Across the Midland region, the resulting degradation values are overwhelmingly negative, commonly ranging from 10 to 30%.
Braden Bowie (Sun,) studied this question.