Developed countries have registration systems that allow near-perfect measurement of death rates, but more than half of the people who die in the world disappear without ever having their existence recorded. Researchers currently lack a reliable strategy for estimating adult death rates in these situations, and many important theoretical and policy questions have thus remained unanswered. I introduce network reporting methods, which combine insights from sociology, demography, and anthropology, to produce survey-based estimates of death rates based on respondents' reports about their personal networks. I show results from the first empirical application of the method in household survey from Rwanda (n=5,000). I also describe an experiment embedded in the Rwanda survey, which serves as a framework for continually improving estimates produced by the new method over time. I conclude with a description of future directions.