This paper provides a systematic overview of the compilation and reliability of national accounts statistics. It illustrates the various issues with a wide range of examples and stories from national accounts compilation practice. National accounts statistics are estimates of a universal accounting model (SNA93). The operational versions of the model decide what is actually estimated. They are estimated by expanding and transforming the available data with accounting identities, assumptions, and plausibility checks. The estimates reflect personal knowledge and skills, resources, and policy. For a specific type of use, the universal and operational national accounting concepts are usually not perfect. The quantitative importance of such conceptual “measurement errors” is often overlooked but can be substantial. For assessing the reliability of national accounts statistics, sampling theory is not very important. The major methods are consistency checks, sensitivity analysis, and analysis using a description of the data sources, operational model, and compilation methods.