Glossary

Comp Database

Also known as: transaction comp database, comps platform, CoStar, CompStak

A subscription platform aggregating sale, lease, and financing transaction data — pricing, terms, parties, and property characteristics — used to support market studies, broker opinions of value, appraisals, and underwriting rent and cap rate assumptions.

Comp databases assemble transaction records from public filings, broker-submitted data, and proprietary research teams to give subscribers searchable, filterable access to recent sale comps, lease comps, and, in more sophisticated platforms, space availability and tenant lease expiration data by submarket and property type. The reliability of a comp database depends heavily on data recency and verification — comps for privately negotiated transactions are sometimes self-reported or estimated, and off-market or portfolio transactions can be mispriced or excluded entirely, meaning experienced analysts corroborate database comps against direct broker relationships and public record searches rather than relying on any single source in isolation. Comp database access and market data licensing has become a meaningful recurring operating cost for brokerage, appraisal, and investment management firms, and is a primary input feeding automated valuation models and AI-powered underwriting tools. The two dominant real-world platforms take philosophically different approaches to sourcing this data: **CoStar** builds and licenses its own proprietary research, employing research staff to verify and maintain listings and transaction records directly, while **CompStak** runs a crowdsourced exchange model, where brokers and appraisers submit their own closed lease comps in exchange for access to comps submitted by others. Neither approach is strictly more accurate — CoStar's model trades submission-network breadth for direct research control, while CompStak's trades proprietary research overhead for real-time practitioner-submitted data — and many firms subscribe to both rather than picking one.