About half of US natural gas comes from gas shales. It is valuable to study field production well by well. We present a field data-driven solution for long-term shale gas production from a horizontal, hydrofractured well far from other wells and reservoir boundaries. Our approach is a hybrid between an unstructured big-data approach and physics-based models. We extend a previous two-parameter scaling theory of shale gas production by adding a third parameter that incorporates gas inflow from the external unstimulated reservoir. This allows us to estimate for the first time the effective permeability of the unstimulated shale and the spacing of fractures in the stimulated region. From an analysis of wells in the Barnett shale, we find that on average stimulation fractures are spaced every 20 m, and the effective permeability of the unstimulated region is 100 nanodarcy. We estimate that over 30 years on production the Barnett wells will produce on average about 20% more gas because of inflow from the outside of the stimulated volume. There is a clear tradeoff between production rate and ultimate recovery in shale gas development. In particular, our work has strong implications for well spacing in infill drilling programs.