In this work, we aimed to generate short- and long-term biological variation estimates of 41 blood metabolites quantified using an NMR-based metabolomics approach. To this end, 30 healthy participants (14 males and 16 females) were recruited and provided serum weekly for 10 weeks and monthly for 10 months, following the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group on Biological Variation recommendations. Creatinine was also measured enzymatically ((IDMS)-traceable enzymatic colorimetric assay) for validation. NMR analytical variation (CVA) was calculated from pooled samples (one pool per dataset) while CVA of enzymatic creatinine assay was assessed with manufacturer's QC material. For between-subject variation (CVG), data normality was initially evaluated with a shapiro-wilk test and non-normal data were log-transformed. Outliers were then assessed at three levels : analytical (Bartlett test), within-subject (Cochran test) and between-subject (Dixon-Q test). For within-subject variation (CVI), individual trends were also evaluated using linear regression and data from individuals not in steady-state were adjusted using the inverse regression formula. After data transformation and outliers removal, CVI (95% CI) was assessed with CV-ANOVA and CVG (95% CI) with classical ANOVA. Raw datasets of NMR concentrations and enzymatic creatinine measurements for both weekly and monthly samples, along with QC variances and all R scripts, are provided here.