CoStar kit: file size and row count table, plus verified join match counts between the JSONL and CSV files
Use commercial-property.id as the hub: apartment-listing.propertyId (1,747 of 1,747 rows), commercial-listing-summary.propertyId (1,022 of 1,022 rows, 629 distinct) and location-occupancy.propertyId (74,093 of 74,093 rows, 2,921 distinct) all match it fully, and commercial-property has 3,339 unique ids. Row counts (wc -l): commercial-property 3,339 (193 MB), apartment-listing 1,747 (6.3 MB), commercial-listing-summary 1,022 (14 MB), commercial-sale-comp 6,120 (23 MB), company-detail 49,575 (42 MB), location-occupancy 74,093 (69 MB), residential-listings 2,120 (45 MB), residential-listing-features 1,899 (28 MB); commercial-properties.csv 13,147 rows (686 KB, 3,339 distinct PropertyId), residential-listings.csv 2,224 rows (150 KB). Counted with python3 -I scripts run outside the kit directory on the local kit files; the non-obvious joins (CSV type mismatch, sale comps, residential, tenants) are covered in separate notes.