In proficiency testing (PT), laboratory performance classifications are used to support conformity assessment and accreditation-related decisions. This study performed an Electrical Materials PT Program and investigated the effect of four commonly used estimators on laboratory performance classifications using the data related to the tensile strength of cable sheathing. To evaluate classification stability, performance categories obtained from the complete dataset were compared with those obtained after removal of two outlying results. Laboratories were classified as satisfactory, warning, or action signal according to their z-scores, and changes in classification were examined for each estimator. The results showed that estimator selection can influence conformity assessment outcomes even when the underlying measurement results remain unchanged. The arithmetic mean exhibited the greatest classification instability, resulting in changes in the performance status of 2 out of 50 laboratories (4%) following outlier removal. Hampel and median estimators each produced one classification change, whereas Algorithm A maintained identical classifications for all laboratories before and after outlier removal, demonstrating the highest classification stability. Furthermore, Hampel showed a slightly more conservative behavior near decision limits, increasing the likelihood of assigning borderline laboratories to the action category. These findings demonstrate that the choice of statistical estimator is not merely a computational consideration but a factor that can affect laboratory classifications and introduce decision uncertainty in proficiency testing. The results highlight classification stability as an important criterion in the selection of statistical methods for PT evaluation and support the use of robust estimators when participant results are close to decision thresholds.