The interpolation of geotechnical properties, such as the N-value in the standard penetration test, plays a crucial role in site characterization and subsurface visualization. However, the sparsity of data often poses challenges in characterizing their spatial distribution and variability. In this study, a Weight Ellipse Nearest Neighbors (WENN) is proposed as a novel interpolation method for sparse geotechnical properties. The method can estimate the spatial mean with standard deviation for uncertainty quantification. The parameters of WENN are data-driven by site-specific measurements using leave-one-borehole-out approach. WENN is applied to a simulated non-stationary SPT example and the effect of number of measurements is discussed in detail. Furthermore, WENN is applied to a field example from New Zealand Geotechnical Database, and its capability is validated by cross-validation tests and comparative results with traditional methods. The outcomes indicate that WENN can capture the overall trend of measurements with acceptable errors, reasonable uncertainty, and sufficient interpretation.