Geostatistical modeling is becoming an increasingly common practice in geotechnical studies. Modern modeling techniques allow advanced interpolation of data and substitution of one piece of information for another, more useful in geotechnical analysis [1]. However, it is not always possible to consider that the picture of the subsoil structure obtained from these methods is statistically controlled information to such degrees as to enable risk analysis. An additional question also remains as to how the reliability of the model is affected by, for example, the frequency of the distribution of survey points, especially with regard to subsoil surveys of linear objects such as roads. The paper presents the capabilities and limitations of a geostatistical ground model prepared for a hypothetical linear facility. Recognition of the properties of the subsoil was made by means of geological surveys and CPTU static soundings. The analysis adopted the actual results of the investigations, obtained in the case of simple and complex geological structure. During the construction of models, data collected at different survey frequencies were used. Techniques that provide statistical control of the homogeneity of geotechnical layers, based only on measured data and not on interpolated or correlated data, were used to create the model. Interpolation and correlation techniques were used only to identify the course of individual layers in the subsoil. Among other things, the results made it possible to assess differences in the statistical reliability of individual model solutions.