The Digital Public Library of America (DPLA) began aggregating data in 2012 and launched its public interface and website in April 2013. That initial set of 2 million records from 16 providers (some of which represented state or community-based aggregations themselves) has since grown to more than 20 million records from 40 providers, who collectively represent around 3,000 individual institutions across the USA. Over the last five years, work on metadata quality at DPLA has shown that to make good decisions about content, coherence and conformance to standards, providers must understand the context of the aggregation with which their records are being shared. This paper reviews the existing literature on metadata quality analysis, and provides an analysis of the metadata quality initiatives at DPLA. DPLA’s work shows that it is more effective to use a combination of automated and community-driven methods to improve data quality than to use either approach in isolation.