Deep cement mixing is widely used to improve soft clayey soils as part of stability intervention in deep excavation and tunnelling. The quality of the improvement is often evaluated by measuring the strength of cored samples post-installation. Although spatial variability in the strength of cement-mixed soil is well recognised, there is still no well-accepted framework for interpreting core strength data. In particular, it is still unclear as to how a representative strength should be derived from the core data. This issue is examined herein using data from two field cases. The core data are first normalised with respect to the duration between treatment and testing, to account for the increase in strength with time. Second, the robustness of two types of criteria is then examined and the sample minima are shown to be relatively sensitive to outlying data. A set of theoretical equations on the variability of sample attributes, such as mean, variance and representative strength, with sample size is then verified using subset re-sampling from the data set, thereby providing a basis for assessing sufficiency of sample size. Finally, a method of conservatively estimating population mean strength and variance from limited sample size based on a prescribed confidence level is also proposed.