Data lakes hold enormous promise in supporting modern enterprise data architectures. Their ability to store a full variety of data types and handle ever-larger data volumes generated at high velocity have contributed to their widespread adoption. But while data lake implementations have technically succeeded in physically consolidating enterprise data, many implementations have failed to deliver returns from a business perspective. That is because much of the data within a lake (and even in more modern lakehouse architectures) remains unconnected and stored in its native form, requiring significant data engineering effort (time and money) to prepare it for analysis.