Cell metabolomics is a valuable tool in oncology. To obtain a valid description of the intracellular metabolome, sample pretreatment is critical to the metabolomics workflow, as this step might influence the reliability of research outcomes. Here we report a systematic investigation to optimize each step of the sample preparation. We explored four quenching protocols, three disruption protocols and seven extraction protocols, and assessed them for effectiveness and robustness by profiling different kinds of representative metabolites by the LC-MS technique using the human glioma cell line U87MG as a model. The optimized pretreatment method for global metabolomics was as follows: after the cells were washed with PBS, cell metabolism quenching was completed within five minutes with 60% methanol (buffered with 0.85% (w/v) AMBIC) at −40 °C. Then, 80% methanol was added and three freeze–thaw cycles were completed to extract metabolites, which could eliminate environmental stress and allow broad coverage of the authentic intracellular metabolome. We also highlighted the advantages and disadvantages of some preanalytical processes for specific metabolites. In addition, the influences of different cell collection batches and liquid nitrogen storage durations were noted. The results suggested the risk of non-physiological or misleading results when using cell samples cultured in different batches or subjected to long-term storage.