Understanding pathways from non-lethal violence to lethal violence between intimate partners is a notable challenge for both policy and practice in partner violence prevention. Of particular interest is whether lethal violence represents an "escalation" of violence from "low" to "high" risk over time, or is best predicted by specific behavioral "typologies" of perpetrators. Testing the "escalation" and "typology" theories is hampered in Australia by limitations in knowledge about non-lethal and lethal violence against women. This article discusses data limitations, measurement problems, and conceptual shortcomings, and suggests approaches to improving evidence quality in the field of violence prevention and risk assessment.