Evaluation of SSOF performance on synthetic datasets. We generate 100 synthetic datasets of HD 26965 (left) or Barnard’s star (right) using the SSOF model and analyze them as if we did not know the parameter values used to generate the synthetic data. This figure reports the percentage of datasets for which SSOF chose to use the same number of feature vectors as was used to generate the synthetic dataset as a function of spectral order. The top panels show results for the number of telluric feature vectors and the bottom panels show results for the number of stellar feature vectors. This demonstrates that SSOF is able to accurately recover the correct number of feature vectors used for the vast majority of orders (excluding those with saturated absorption features) for realistic survey parameters (here, 100 observations each with typical spectral SNR).