BigQuery cost optimization starts with reducing the data processed by each query and matching capacity to workload demand.
Partitioning lets BigQuery skip partitions that a query does not need. Clustering organizes data within partitions so filters can reduce the blocks read. Queries that filter on partitioning and clustering columns generally process less data than equivalent queries without those structures.
On-demand cost follows bytes processed, while Editions cost follows capacity and slot usage. Partitioning can reduce on-demand bytes and also shorten execution on capacity-based workloads. Clustering has its own maintenance and storage considerations.
Measure bytes processed, slot time, runtime, and storage before and after a table change. A small query test does not represent a production workload with different filters or data volume. Google documents partitioned tables, clustered tables, and pricing.