Purpose: This study examines the usefulness of a behavior-based typology in the prediction of completion of batterer intervention program (BIP) and reoffense. It additionally explores typology differences by class and ethnicity. Method: 899 offenders completed a 2-hour psychosocial assessment and were referred to one of 31 BIPs. Cluster analysis determines whether a unique typology exists as indicated in a previous analysis of a subsample from the same population of offenders (n=158). The predictive ability of the typology on BIP completion and reoffense is tested using completion and re-arrest data. Cluster analysis by ethnicity and class will further explore variations in the typology. Results: Previous analysis of the smaller subsample provided evidence of a behavior-based tri-fold typology (Low-level criminality, Dysphoric volatile, and Dysphoric Generally Violent). Preliminary analysis of the predictive ability of the typology on program completion within the subsample indicates the typology is able to identify men likely to require multiple referrals to BIP before completing. Dysphoric Volatile and Dysphoric Generally Violent men were more than 2 times likely (25% compared to 11%) to need multiple referrals to a BIP before completing. Conclusions: This behavior-based typology may have better predictive ability than previously used personality-based typologies. Indications are that a Coordinated Community Response (multiple referrals) with men previously shown to be most likely not to complete were held accountable and eventually did complete.