I apply a regression-kriging model to production and engineering data from thousands of tight oil wells to demonstrate that the role of technological learning has been overestimated by current, less accurate methods which fail to account for "sweet-spotting," or shifts of drilling activity toward the best acreage.