A system and method for identifying causal email threading. In one aspect, a computing device identifies a plurality of email templates, each email template corresponding to characteristics of a received machine-generated email, the characteristics of the received machine-generated email relating to static data of the machine-generated email. The computing device generates a template causality graph by analyzing the plurality of email templates to determine a statistical causality between templates of the plurality of email templates, the determining of the statistical causality between templates comprising determining that a first received machine-generated email associated with a first template is a result of a second received machine-generated email associated with a second template.