The interest in marine renewable energy is strong, but has not led to significant commercial-scale investment and deployment, yet. To attract investors and promote the development of a marine renewable industry a clear concept of project risk is paramount, in particular issues relating to device reliability are critical. In the public domain, reliability information is often scarce or inappropriate at this early stage of development, as little operational experience has been gained. Thus, reliability estimates are fraught with large uncertainties. This paper explores sources and magnitudes of failure rate uncertainty and demonstrates the effect on reliability estimates for a notional marine energy converter. If generic failure rate data forms the basis of a reliability assessment, reliability estimates are not robust and may significantly over- or underestimate system reliability. The Bayesian statistical framework provides a method to overcome this issue. Generic data can be updated with more specific information that could not be statistically incorporated otherwise. It is proposed that adopting such an approach at an early stage in an iterative process will lead to an improved rate of certainty. (C) 2012 Elsevier Ltd. All rights reserved.