The cloud data platform combines capabilities of data warehouses and data lakes to support analytics, requiring flexibility in data access, performance, and ease of transformation. To meet these requirements, data teams need a platform that structures data in multiple ways, creating and applying schemas at different points in the data lifecycle depending on use cases. Two traditional approaches are schema on write and schema on read, each with its pros and cons: schema on write provides flexibility but reduces performance and creates upfront cost, while schema on read offers flexibility but slows query performance and requires complex effort. A new approach is proposed by Thomas Hazel, leveraging virtual schemas and compressed indexing to balance flexibility and performance.