The human immune system is known to be highly variable among individuals, but it is not well understood how the variability changes over time, especially when faced with external perturbations. Here we analyzed individual variability in the immune system in a cohort of 301 Japanese volunteers who received the same trivalent inactivated influenza vaccine in winter 2011. To extract important variability axes from single-cell measurements in a data-driven and unsupervised manner, we devised a computational method termed LAVENDER (latent axes visualization and evaluation by nonparametric density estimation and multidimensional scaling reconstruction). It measures distances between samples using k-nearest neighbor density estimation and Jensen-Shannon divergence, then reconstructs samples in a new coordinate space, whose axes can be compared with other omics measurements to find biological information. Application of LAVENDER to multidimensional flow cytometry datasets of B and T lymphocytes (taken before and 1, 7, 90 days after vaccination) uncovered an axis related to time and another axis related to individuality. We found that the values of the individuality axis were positively correlated between different days, suggesting that the axis reflects the baseline immunological characteristics of each individual. In fact, the value of the axis before vaccination was highly correlated with the neutrophil-to-lymphocyte ratio, a clinical marker of the systemic inflammatory response; this was verified by the transcriptome analysis of peripheral blood. These results demonstrate that LAVENDER is a useful tool for identifying critical heterogeneity among similar but different single-cell datasets.