<p>This paper develops a spatial autoregressive stochastic frontier model that allows the spatial weighting matrix to be stochastic. This approach captures spatial heterogeneity in the interactions among production units and quantifies random spatial dependencies arising from this variability. The model parameters are estimated using a simulated maximum likelihood approach. We propose a method to simulate spillover-corrected direct, indirect, and total efficiency estimates. Additionally, we conduct Monte Carlo simulations to evaluate the finite-sample performance of the proposed model. Finally, an empirical application to the efficiency of Spanish dairy farms is presented to demonstrate the practical implementation of our methodology.</p>