The ages of critical periods in the geological record of the Earth have to be determined and attached uncertainties quantified as precisely and accurately as possible. This study tries to find an optimal strategy for sampling, for both biochronology and high-precision U–Pb analysis of zircon fromash beds, and for the selection and statistical treatment of geochronological data. This strategy is applied to the case of the Early–Middle Triassic boundary (EMTB) in the southern Nanpanjiang Basin in South China, aiming to obtain the most accurate age information for this boundary with an associated uncertainty. The EMTB has been located in a sedimentary section near Monggan, Guangxi, South China, which permitted tight correlation between conodont biozones, the carbon isotopic record, and high-precision zircon U–Pb data from volcanic ashes. Two residual maximal horizons were defined using characteristic associations of conodont species, accurately bracketing the EMTB. The lower and upper age limits of these horizons, as well as the first occurrence (FO) of the base Middle Triassic index fossil Chiosella timorensis, were numerically calibrated by high-precision chemical abrasion, isotope dilution, thermal ionization mass spectrometry (CA-ID-TIMS) U–Pb dates. Volcanic zircon in the analyzed ash beds reveals complex system behavior, in particular (i) unresolved lead loss after chemical abrasion, and (ii) pre-eruptive growth and/or recycling of zircon crystals, leading to excess scatter in 206Pb/238U dates and to violations of the stratigraphic sequence. A sufficiently high number of closely spaced ash samples (N) and of U–Pb analyses per sample (n) promotes recognition of such complications. Two different age models were constructed from weighted mean 206Pb/238U dates, comparing smoothing cubic spline and Bayesian interpolation approaches. The statistical treatment of the combined datasets allows quantification of the lower and upper limits and associated uncertainties of the two conodont biozones as well as of the FO of C. timorensis. The age for the EMTB is calculated from combined interpolated data at 247.05 ± 0.16 Ma. The age model derived from Bayesian interpolation is considered to be more adapted to the present case, where discontinuous changes in sedimentation rates are expected to be important. The smoothing spline represents a more parsimonious model, producing a simpler solution in terms of smoothness, but the model assumptions are too strong for many geological problems. The additional freedom in constructing the Bayesian interpolating curve presumably leads to more realistic, albeit larger confidence intervals.