To accurately estimate per turbine – or per megawatt – annual wildlife mortality at wind facilities, the raw counts of carcasses found must be adjusted for four major sources of imperfect detection: (1) fatalities that occur outside the monitoring period; (2) carcasses that land outside the monitored area; (3) carcasses that are removed by scavengers or deteriorate beyond recognition prior to detection; and (4) carcasses that remain undiscovered by searchers even when present. To accurately estimate regional or national annual wildlife mortality, data must come from a representative (or appropriately weighted) sample of facilities for which estimates of mortality account for all sources of imperfect detection. I argue that the currently available data in the United States and much of the world do not represent the impacts of wind power on wildlife because not all facilities conduct monitoring studies, not all study results are publicly available, and few studies adequately account for imperfect detection. I present examples illustrating the limitations of our current data and pitfalls of interpreting data without accurately adjusting for detection bias. I close by proposing a solution through a simplified monitoring process that can be applied at every facility as part of normal operations. Application of an unbiased estimator that accounts for all sources of imperfect detection would assure comparability of mortality estimates. Public access to reported estimates would achieve representation. With these data we could develop a clearer understanding of how wind power is affecting wildlife throughout the world and inform our efforts to address it.