Compensation specialists made two survey-sample decisions in a simulated wage survey. Policy-capturing analyses indicated that most specialists relied extensively on two of the available cues and consistently applied that policy across judgments. They were not able to accurately estimate their own decision-making policies, as demonstrated by the fact that rationally generated assessments of cue importance were significantly different from their actual policies. Finally, meta-analyses demonstrated that the variation in policies across decision makers could not be attributed to statistical artifacts or any moderators (e.g., salary-survey experience or industry) associated with the demographic data. Thus, different compensation specialists are likely to select a different sample of employers even for the same wage survey. Implications for the relevance of market wages obtained from surveys are discussed.