Two studies explored triage decision-making in a welfare fraud investigation, specifically decisions concerning what evidence to collect when deciding whether to pursue a case to prosecution or just to issue a warning. An observational study revealed that triage decisions appear to be determined by subjective estimates of the ease of evidence collection, and that these estimates are influenced by complexity of mapping evidence onto fraud types. This hypothesis was explored in an experimental study of investigators, managers, and students choosing evidence to inform triage decisions for cases that varied according relevance and complexity. Students’ selections were unaffected by the nature of the case. In contrast, with a simple fraud case, investigators and managers tended to select evidence to support a prosecution decision, but with a complex fraud case they selected evidence that supported comparison of prosecution and warning decisions. The results demonstrate flexible expertise in choosing what evidence to sample.