Cone penetration testing (CPT) has become one of the most commonly adopted in situ tests for site investigations, due to its reliability and repeatability of the measurements and the potential for direct use in design. For CPTs in clays, previous studies have paid considerable attention to the selection of penetration resistance factor N-kt considering the effects of soil stiffness, stress and strength anisotropy, soil layering and soil sensitivity, and so on. In contrast, this paper focuses on the effects of spatially variable soil properties on the interpretation and use of CPT data. This has been achieved using large-deformation finite-element modelling with random fields. It is shown that when using CPT data to obtain the true point-to-point statistics of the ground, scale effects could lead to the (unconservative) overestimation of the low estimate and underestimate of the high estimate for variable ground with small scales of fluctuation. Suggestions are made for correcting CPT-measured data to allow for this effect. In contrast, if using CPT data to provide design inputs for (larger) foundations, the same scale effects may be considered to increase beneficially the characteristic values of soil strength for sizing and reduce characteristic soil strengths for installation assessment.