Although there has been considerable research into the effectiveness of individual cognitive behavioral treatment for chronic insomnia, less is known about patients' perceptions of what constitutes actual improvement. This study utilized 70 outpatients (mean age = 49.7 years, SD = 12.0) with insomnia who completed a 6-week cognitive behavioral group for sleep. Participants completed a number of primary (Pittsburgh Sleep Quality Index) and secondary measures (the Dysfunctional Beliefs about Sleep Scale, Insomnia Severity Index, Beck Depression Inventory, Penn State Worry Questionnaire) at pre- and post-treatment. Perceived improvement was measured using the Clinical Global Improvement Scale (CGI). Results were analyzed using a combination of Logistic Regression analysis and receiver operating curve characteristic analysis (ROC). Results demonstrated that sleep quality and sleep duration were the most sensitive primary measures, or best predicted perceived improvement, whereas sleep efficiency was the most specific primary measure, or best predicted perceived lack of improvement (defined as only mild improvement). Of the secondary measures, results showed that daytime impairment was the most sensitive predictor of perceived improvement and that mood was the most specific predictor of perceived lack of improvement. Implications of these findings are that sleep quality, sleep duration, and sleep efficiency may offer different types of information and the choice of measure for predicting global improvement in insomnia will depend on the needs of the researcher/clinician.