Abstract
It is not a secret that well productivity within the US Land part of the oil & gas industry has been subject to significant improvements in the recent years. Official state data suggests that this process was accompanied by a steep growth in typical well configuration and completion intensity metrics (e.g. lateral length, proppant and frac fluid intensity, number of frac stages per well). In addition to this shift towards more complex wells, a “high-grading” of drilling locations was happening behind the scenes. The importance of high-grading for well productivity improvements can be difficult to quantify. Another arguably important driver of shale well productivity is the order of completion round for the considered pad or lease. The main purpose of this work is to build a comprehensive statistical model which reveals actual impact of well design, acreage quality and child wells on shale well productivity. One of the most prolific unconventional liquid plays, Permian Delaware, is utilized as a case example.
Several regression models are constructed with thorough assessment of embedded multicollinearity problems as many independent well design parameters exhibit strong correlation with each other. Cumulative total two-stream production over different time initial production periods is used as a dependent variable in all models. We consider horizontal wells targeting one of the Wolfcamp or Bone Spring benches and exclude all wells with missing lateral length or proppant mass.
We discover that simultaneous inclusion of proppant and frac fluid variables into the model results into misleading interpretation and significance of individual coefficient estimates amid strong collinearity. Therefore, we exclude frac fluid from the final versions of the model as its predictive power is largely carried by proppant.
When controlling for spatial terms (acreage quality), predictive power of the models jumps from 35-40% to 45-50% in terms of explained variability with the relative predictive power of acreage quality increasing from less than 20% for 90-day cumulative production to 35% for 2-year cumulative production. We find strong empiric evidence that the impact of rock quality on well recovery rates increases as well gets older.
We find strong statistically significant negative impact of child wells on well productivity.
We present an innovative approach to build initial expected productivity estimates for different landing zones in the Permian Delaware. These estimates could be utilized by operators as a starting point when exploring new parts of the basin.
Our models allow to develop reasonable and justified productivity normalization metrics. While linear normalization of productivity per perforated lateral foot or even per pound of proppant pumped are frequently used metrics, we provide evidence suggesting that such metrics should be used with caution.