Xylogenesis, the process through which wood cells are formed, results in the long-term storage of carbon in woody biomass, making it a key component of the global carbon cycle. Understanding how environmental drivers influence xylogenesis during the growing season is therefore of great interest. However, studying short-term drivers of wood production using xylogenetic data is complicated by the usual sampling scheme and the influence of eccentric growth, i.e., heterogeneous growth around the stem. In this study, we improve xylogenesis research by introducing a statistical approach that explicitly considers seasonal phenology, short-term growth rates, and growth eccentricity. To this end, we developed Bayesian models of xylogenesis and compared them with a conventional method based on the use of Gompertz functions. Our results show that eccentricity generated high temporal autocorrelation between successive samples, and that explicitly taking it into account improved both the representativeness of phenology and intra-ring variability. We observed consistent short-term patterns in the model residuals, suggesting the influence of an unaccounted-for environmental variable on cell production. The proposed models offer several advantages over traditional methods, including robust confidence intervals around predictions, consistency with phenology, and reduced sensitivity to extreme observations at the end of the growing season, often linked to eccentric growth. These models also provide a benchmark for mechanistic testing of short-term drivers of wood formation. • Eccentric tree growth biases xylogenesis monitoring. • Bayesian models outperform Gompertz for interpreting xylogenesis data. • They may explicitly account for phenology, eccentricity, and intra-annual forcing. • Gompertz-based methods require caution in application.