Histogram showing the frequency distribution of model per cent rms errors for random experiments of 28 observations executed as second-stage surveys, compared with an adapted experiment also executed as a second-stage survey (black line). The data were noiseless. The input model for all inversions was the stage-one dipole-dipole inversion in Fig. 5. The black line at left shows the model per cent rms error attained for the adaptively designed experiment of 28 observations. The model per cent rms error of the designed experiment is in the lowest 5th percentile, meaning it outperformed 95% of the random experiments.