PostgreSQL estimates selectivity using statistics about table values. ANALYZE refreshes those statistics. A plan can choose poorly when the source changed substantially but its statistics still describe the old distribution. Compare estimated output rows with actual output rows before blaming the scan method.
A column statistics target controls how much detail ANALYZE collects. Increasing it can improve estimates for uneven distributions, at a cost in collection and planning work. Extended statistics describe relationships between columns, such as correlated tenant and status values, that independent single-column estimates can miss.
Statistics don't create an index and can't guarantee a particular plan. These exercises inspect the target and stored statistics definitions directly. The data is small enough for repeatable checks, so its runtime doesn't establish a production tuning benefit. Collect representative evidence before changing global planner settings.
ANALYZE updates the table row estimate rather than counting only populated amounts.