In the quality assurance procedure of the deep cement mixing method, the statistical parameters of the unconfined compressive strength quf of core samples are adopted to assess the quality of the improved ground. Since the mean and the standard deviation of core strength are the sample statistical parameters, the statistical uncertainty emerges when estimating the population mean and standard deviation. Moreover, the spatial correlation of quf affects the overall performance of cement-treated soil columns. The finite element method with random field theory, random finite element method (RFEM), is a useful tool for evaluating the influence of the spatial correlation of quf. Based on the probability characteristics provided by the RFEM analysis, a reliability-based assessment is possible in the quality assurance procedures. The paper presents a reliability-based assessment for the deep mixing cement soil column based on the core strength. The author has proposed the analysis method in which the statistical uncertainty included in the core strength data and the spatial variability of the strength are considered simultaneously (Namikawa 2021). In the proposed analysis method, the realizations of the statistical uncertainty are estimated using a Bayesian inference method and the random fields of strength are generated with the realizations involving the statistical uncertainty. The statistical parameters of the strength were estimated using the data acquired in practical projects. The estimated statistical parameters of quf were adopted to generate the random fields of strength for a cement-treated soil column model. In the RFEM analysis, the compression behavior of a cement-treated soil column was simulated to calculate the overall strength Quf of a cement-treated soil column. The analysis results provide the probability distribution of Quf of the column. Based on the probability distribution, the reliability-based assessment was performed by calculating the failure probability for the design strength.