Recent years have witnessed the flourishing of literatures dealing with electoral integrity and forensic techniques by which electoral fraud can be identified and analysed. Statistical techniques are now available through which highly improbable patterns of voting can be highlighted as part of the process of investigating whether an election deserves to be categorised as free and fair. Yet relatively little use has been made on the ground of such forensic techniques by the international community when fraud is suspected. By contrast, in Afghanistan in 2009, a 'sampling' process was deployed to investigate suspicions of fraud; and in 2014, a box-by-box 'audit' of ballot boxes was carried out. At the end of the day, significant groups were left deeply dissatisfied with the ways in which suspected fraud had been investigated; and such dissatisfaction could potentially be the basis for further elite fragmentation and a deepening legitimacy crisis. It is argued in this article that it is important to ensure that mechanisms for investigating fraud be defensible not just in terms of what has been done in the past, but in terms also of what is methodologically rigorous and defensible.