A data engineering team maintains a reporting dashboard that visualizes transactional records from Firestore in Datastore mode. Users report severe latency degradation when navigating to deeper pages in the data grid. An audit reveals that the underlying application issues queries that retrieve all 50 entity properties while using integer OFFSET and LIMIT clauses for pagination. Which combination of data modeling and query design changes should you implement to resolve the latency issues and reduce operational read costs?

Replace the query offset with query cursors for pagination, and rewrite the query as a projection query that specifies only the required dashboard properties.: ### What This Solution Is This solution combines **query cursors** with **projection queries** in Firestore in Datastore mode to eliminate internal read amplification and reduce payload overhead. ### How It Meets the Requirements - **Remediates latency degradation during pagination**: Standard `OFFSET` clauses force the database engine to fetch and evaluate all skipped entities internally before discarding them, leading to linear $O(N)$ latency growth on deeper pages. **Query cursors** act as bookmark tokens indicating the exact index position where the next page begins, executing in $O(1)$ constant time regardless of page depth. - **Reduces operational read costs**: When using `OFFSET`, your application is billed for reading every skipped entity. Cursors ensure you are only billed for the specific entities returned on the active page. - **Optimizes data transfer and serialization**: Rather than retrieving all 50 entity properties across the network, a **projection query** fetches only the specific subset of indexed properties required by the visualization component, minimizing deserialization latency and memory consumption. ### Key Advantages and Capabilities - **Constant-time pagination**: Cursors provide consistent response times across thousands of paginated records. - **Reduced index scan overhead**: Projection queries read values directly from index entries without fetching the underlying complete entity. - **Direct cost reduction**: Eliminates billed read operations for discarded offset records. - **Minimal network throughput**: Significantly reduces payload sizes returned to the visualization layer. ### Why This Is Optimal Using cursors alongside projection queries directly targets the two primary architectural bottlenecks: unindexed pagination scanning and over-fetching of unneeded entity properties.