This paper presents a concept and experimental results for a novel confidence metric for speaker identification (SID). The concept is based on three major audio components: duration, signal-to-noise ratio, and model quality. This concept has been successfully applied to two benchmark closed-set SID systems which demonstrates the broad applicability of the technique. A subset of the TIMIT database was used for this investigation. Derivation and validation of the con- fidence performance are reported in this paper. We also investigate a fusion technique for the composite multi-component confidence metric. Reliability of the confidence metric has been determined to be 94.6% accurate on average.