Accurate determination of intracellular metabolite levels requires reliable, reproducible techniques for sampling and sample treatment. Quenching in 60% (v/v) methanol at ?40°C is currently the standard method for sub-second arrest of metabolic activity in microbial metabolomics but there have been contradictory reports in the literature on whether leakage of metabolites from the cells occurs. We have re-evaluated this method in S. cerevisiae using a comprehensive, strictly quantitative approach. By determining the levels of a large range of metabolites in different sample fractions and establishing mass balances we could trace their fate during the quenching procedure and confirm that leakage of metabolites from yeast cells does occur during conventional cold methanol quenching, to such an extent that the levels of most metabolites have been previously underestimated by at least twofold. In addition, we found that the extent of leakage depends on the time of exposure, the temperature and the properties of the methanol solutions. Using the mass balance approach we could study the effect of different quenching conditions and demonstrate that leakage can be entirely prevented by quenching in pure methanol at ??40°C, which we propose as a new improved method. Making use of improved data on intracellular metabolite levels we also re-evaluated the need of sub-second quenching of metabolic activity and of removing the extracellular medium. Our findings have serious implications for quantitative metabolomics-based fields such as non-stationary 13C flux analysis, in vivo kinetic modeling and thermodynamic network analysis.