Every year, an increasingly larger fraction of drilling is done on pads or in zones where wells are already present. These existing wells have implications for the hydrocarbon production of any new developments, and therefore also the economic value of engaging in this activity. Here, we use a non-linear and multivariate machine learning approach to provide descriptive evidence of the effects of existing well production on infill wells and segregate that impact into the contributions of individual features. We find that the percentage of total reserves produced by existing wells before an infill well is brought online is the single strongest factor in determining the relative performance of the infill well as compared to the parent. Distance to the closest parent produces a nonlinear effect on child production, and one which is mediated by the percent of reserves already generated by the parent well. Finally, we observe mixed results for the influence of geology, which warrants further investigation.
As American shale plays mature, operators have re-visited previously drilled locations to add more wells. In the Midland basin, for example, the proportion of "infill" wells has risen from 50% to 75% of new drilling between 2015 and 2021. Part of the motivation behind this change is the increasing price of acreage in American onshore basins, and especially recently, the increase in M&A activity. Another motivation is that tier 1 acreage may represent a lower risk profile for re-drilling compared to new development on lower quality acreage in the basin, or at least a more easily quantifiable risk profile.
While operators know infill development will continue to increase, it has been historically difficult to make informed decisions about the tradeoffs between drilling and completion costs and the potential for production uplift. This is largely because the explanations of why a particular child well produces more or less than their immediate parents have remained largely undetermined.