The inversion of cross-hole data leading to tomographic imaging magnifies model and measurement errors. This effect couples with limited illumination angles and mixed-determination to render irrelevant images which often resemble the spatial coverage of the measurements. Yet, proper inversions may still be obtained by adding information to the inverse problem. This paper centers on the pre-processing of cross-hole tomographic data to identify the general characteristics of the host medium, the presence of anomalies, and the overall nature of the inverse problem. Several pre-processing strategies are examined within the context of two well-documented case histories. Results confirm the ability of pre-processing to provide foresight about the medium to be imaged and to help select an adequate inversion strategy.