Recent studies have shown that providing feedback to therapists based on comparing their clients' progress to a set of rational, clinically derived algorithms delineating various categories of progress has enhanced outcomes for clients predicted to show poor treatment outcomes (signal-alarms). One prior study indicated that an empirically derived method more accurately predicts outcome than the rational method. The present study replicated the comparison of empirical and rational methods, while adding an additional layer of effect size analyses to further clarify predictive accuracy. The two methods were approximately equivalent in their accurate detection of cases that had a final negative outcome. However, the rational method had significantly lower overall predictive accuracy due to its high percentage of false negative predictions. Further, progressively more optimistic predictions based on the empirical method were associated with greater improvement for the average client. This was not the case for the rational method. Copyright © 2006 John Wiley & Sons, Ltd.