How BrokerHQ Builds Tenant Intelligence for CRE

    Updated August 23, 2026

    The CRE data industry has a trust problem: opaque sources, stale records, and "intelligence" that's a polite word for guesswork. BrokerHQ exists because brokers deserve data they can show to a principal without flinching.

    This page documents how we build our intelligence layer: where the data comes from, how we verify it, how often we refresh it, and what we don't do. If anything on a BrokerHQ profile doesn't add up, you should be able to trace it back to a source from this page.

    Our data sources

    We aggregate from four primary streams:

    1. Public SEC filings. Every 10-K, 10-Q, 8-K, S-1, and proxy filed by a US public company runs through our ingestion within 24 hours of release. We parse for lease commitments, real-estate footprint disclosures, restructuring announcements, and material-event events that drive CRE decisions.
    2. Structured lease and property disclosures. Public lease records from county registries and municipal property records. This is county-level and city-level data, so its reach is the set of jurisdictions ingested rather than a national figure: today the Puget Sound region, led by King County Assessor and Seattle Department of Construction and Inspections records.
    3. News and trade press. Major business publications, regional CRE trade press, and trusted local business journals. Articles are processed for company-level events relevant to real-estate decisions: expansions, contractions, fundraising, leadership changes.
    4. Proprietary signal streams. Internal models that combine the above with employment, hiring, and corporate-action data to produce composite signals (e.g., a "lease-decision-imminent" score).

    Every record on a BrokerHQ profile links back to its underlying source. If you can't trace a claim, it doesn't ship.

    Refresh cadence

    • Public filings: Within 24 hours of release.
    • News and trade press: Daily.
    • County property records: Weekly where the registry supports it; monthly otherwise.
    • Composite signals: Recomputed nightly.
    • Company profiles overall: Refreshed at least weekly; high-signal companies refresh daily.

    We surface the "last verified" timestamp on every data point so brokers can judge freshness at a glance.

    Verification

    For high-stakes data points (lease maturity dates, specific addresses, executive names), we apply a two-source rule: a claim must be supported by at least two independent sources before it appears in a customer-facing profile. Where two-source isn't possible (e.g., a unique 8-K disclosure), the data point is tagged "single-source" and surfaced with that caveat.

    We do not fabricate. We do not "fill in" missing data with inference. When a record is incomplete, it's labeled incomplete, not invented.

    Measured accuracy

    Our company matching is scored against a golden set of 95 human-verified cases. Current measured performance: 0.975 precision, meaning when we link a record to a company we are right 97.5 percent of the time, and 0.886 recall, meaning we catch 88.6 percent of the links that exist. We publish these numbers so you can judge where the system is reliable. Records that do not meet our verification standard are labeled, not guessed.

    Coverage and known limits

    • Company coverage, national: every US public company that files with the SEC, wherever it operates. Full for public filers; substantial for private companies with public disclosures such as Form D, federal awards or WARN notices; signal-only for private companies with no public filings.
    • Property and local-records coverage, Puget Sound today: King County Assessor parcels and verified sales, Seattle permits and energy benchmarking, Washington SEPA, and Seattle and Washington business licences. Further metros are the expansion path, added one jurisdiction at a time as each publishes usable records.
    • Asset class coverage: Office, industrial, retail, life sciences, flex. Multifamily and hospitality are on the 2026 roadmap and are not currently in production.
    • What we don't cover: Residential real estate. International markets. Off-shore corporate structures with no US filings. Property-level depth outside the Puget Sound region.

    If a market or asset class you need isn't listed, contact us and we'll tell you our roadmap honestly.

    AI and human review

    We use machine learning models for entity resolution (linking a filing back to a known company), event extraction (parsing "we leased 50,000 sq ft in Plano, TX" out of paragraph 14 of an 8-K), and signal scoring. Human reviewers spot-check model outputs on a sampled basis and adjudicate edge cases.

    Where AI assists, we say so. Where a number is computed (rather than extracted), it's labeled as a model output.

    What we don't do

    • We do not buy data from brokers or list-rental aggregators.
    • We do not infer lease terms from "comparable" buildings; if we don't have the actual lease record, we don't claim a lease.
    • We do not auto-generate company summaries with LLMs unsupervised; every summary is template-driven and source-cited.
    • We do not show "estimated employee counts" presented as actual employee counts. Estimated values are labeled estimated.

    Questions or corrections

    If you find a data error on a BrokerHQ profile, email data@brokerhq.ai with the company name and the specific claim. We acknowledge within 24 hours and correct within 5 business days. Our correction log is internal but available on request to enterprise customers.

    Start prospecting in under 10 minutes.

    30-day free trial. No credit card. No demo required.

    Start Free Trial See how it works →