Bitcoin introduced the blockchain as a decentralized data structure to establish consensus about who owns which coins. Since then, blockchain technology has evolved and now enables distrusting parties to engage in various online interactions without trusted intermediaries. However, this technology is not without its own shortcomings. While prior work has extensively studied technical challenges, it neglected the influence of the data stored on the blockchain so far. In this dissertation, we take a data-driven perspective to assess and improve permissionless blockchains as building blocks for decentralized data management systems. We identify two core challenges, i.e., the need for moderating what data is recorded and the need for alleviating the storage requirements of ever-growing blockchains, and assess the technology's potential to foster further applications. With the four contributions we present in this dissertation, we shed new light on the potential impact of the data persisted on blockchains. Our analyses and technical contributions therefore widen the scope for resilient and durable blockchain designs for data management tasks.