Abstract We present a hierarchical model for survey comparative fishing (CF) experiments (x) to utilize data from several species (s) and x to provide improved estimates of the relative efficiency of one survey protocol compared to another. This model is applied to four flatfish s and two x conducted by Fisheries and Oceans Canada (DFO) in 1995 and 1996. We used a monotone increasing function for relative efficiency, and included spatial effects to account for this important source of variation that was not considered in previous analyses of these data. We provide detailed analyses of the anticipated impacts of the various changes in the DFO survey protocols to better understand the reliability of the results. We show that there were important differences in relative efficiency among s, x, and spatial regions, which, combined with low sample sizes and low catch rates, contributed to poor precision in the estimates of relative efficiency. We conclude that stock assessment models in the future should have a goal of using unconverted survey indices, but also include information on the relative efficiency of trawl survey protocols as prior distributions. This will more adequately account for this important source of uncertainty.