How Lupa Builds Parcel Audits on Public Records

Every Lupa finding is independently verifiable against a government source — that's the design constraint. This page documents how data is sourced, how owners are matched across jurisdictions, and how leads are delivered so your audit team can verify before acting.

Data sources

Lupa uses only public records published by government bodies. No paid aggregators (Regrid, ATTOM, PropStream, Melissa Data) are in the pipeline.

• County property appraiser portals — parcel rolls, exemption flags, ownership records.
• State cadastral databases — statewide parcel datasets published by state revenue and property-tax agencies.
• Recorder/clerk offices — deed transfers and title history.
• Permit systems — county and municipal permit histories.
• Aerial imagery — NAIP, county-flown imagery programs, and other openly published sources.
• Business registries — secretary of state filings for entity-owned parcel cross-checks.

Signals

The kinds of evidence Lupa looks at when generating a lead. No single signal becomes a lead on its own — every delivered case requires corroboration across at least two independent primary sources.

• Simultaneous primary-residence claims across jurisdictional boundaries (county-to-county within a state, or state-to-state).
• Mailing-address inconsistencies between parcel ownership records and active exemption or owner-occupied flags.
• Tenure gaps — exemption flags persisting on parcels that have changed hands in ways the receiving record hasn't yet reflected.
• Entity-ownership structures incompatible with primary-residence statutes — trust, LLC, or business-entity ownership where residency law requires a natural person.
• Public-record contradictions across primary-source databases — assessor data, auditor data, recorder filings, and business registries that name the same subject in mutually exclusive ways.

Guardrails

False-positive filters Lupa applies before any lead leaves the pipeline.

• Recent-sale filter — flagged cases with ownership transfers in the last 24 months are excluded as a known false-positive class (exemption flags lag deed transfers in most jurisdictions).
• Common-name disambiguation — owners whose names appear in many parcel records statewide require additional corroborating signals before inclusion.
• Active-duty military exclusion — APO/FPO addresses are excluded under the relevant statutory protections (e.g., FL §196.061).
• Federal-entity exclusion — parcels held by federal agencies, mortgage servicers (Fannie Mae, Freddie Mac), or other sovereign-immunity entities are excluded as uncollectable.
• Confidential-address exclusion — protected-class records (judicial, law enforcement, victim-advocate) under public-records-redaction statutes are excluded.
• Side-by-side review — likely or ambiguous cases require manual side-by-side comparison of source documents before delivery.

Scope discipline

How Lupa interacts with the public data sources it uses.

• Identifiable user agents on every public-API access — server logs attribute scans to Lupa, not to anonymous traffic.
• Conservative request rates with explicit sleep cadence between calls — public APIs are not stress-tested.
• No silent re-scraping after first contact with a target jurisdiction. If Lupa has reached out to a county, that county's data sources are not re-pulled until the conversation has been resolved or 30 days have passed, whichever comes first.

Citation-backed leads

Every lead delivered by Lupa includes:

• Parcel IDs in every relevant jurisdiction.
• Direct URL citations to the source public record (county portal, state cadastral record, permit system).
• The specific data point that triggered the flag (exemption status, sale date, permit gap).
• Confidence tier so audit teams can prioritize.

This means a Lupa lead is not "trust our model." It's a pointer to a public record your team can open and read directly.

Verification workflow

Lupa is an investigative-lead generator, not an enforcement tool. The intended workflow:

1. Lupa delivers ranked leads with citations.
2. Your audit team verifies each lead against the cited public records.
3. Verified cases enter your standard notice/dispute process.

Side-by-side record comparison is the default for likely or ambiguous cases — never act on a flag without confirming the underlying documents.

Known constraints and caveats

• New York STAR program (post-2019) — county roll data alone misses post-2019 STAR credit recipients because the program shifted to a state-administered credit. Reliable NY checks require the NYS state dataset.
• Florida §193.092 — back-assessment recovery on permitted parcels is statutorily capped, so revenue projections on permitted-but-unassessed pool detections must account for this.
• Recent sales — homestead-exemption flags lag transfers, so flagged cases on parcels with recent ownership changes are a known false-positive class and are reviewed before delivery.
• Owner-name conventions are jurisdictional — new states require validation of the local naming convention before adapter coverage is claimed as production-ready.

Why public records only

Three reasons:

• Verifiability — every finding is reproducible by the customer using the same source data, which matters for due process when notices go out.
• Cost structure — no per-record or per-query pricing means scans scale to full-jurisdiction without surprise costs.
• FOIA-friendly — when a flagged owner asks "where did this come from?" the answer is a public URL, not a vendor's proprietary score.

Homestead Exemption Audit →  ·  Pool Detection →