How the screen works, and what it measured.

The test — Tax Code §41.43(b)(3) / §42.26(a)(3)

A property is unequally appraised when its value exceeds the median appraised value of a reasonable number of comparable properties, appropriately adjusted. The screen applies that statutory test, literally, to every improved commercial parcel on the certified roll with five or more comparables (7,741 of 17,418).

Comparables come from the same roll with the same improvement-type code and the same TCAD market area, building size within ±50%, year built within ±15 years, land share within ±0.25 of the subject. Parcels that are mostly land (land share > 60%) are excluded from both sides — land plays need a land screen. Each comparable's improvement value per square foot is adjusted for size and age, and the subject is compared with the median of the adjusted comps.

When a parcel is flagged

Excess over the median ≥ 10% (a business filter — the statute itself has no threshold) in a cohort of ≥ 5 comparables with dispersion (IQR ÷ median) ≤ 0.60. HIGH confidence: ≥ 8 comps, dispersion ≤ 0.25. MEDIUM: ≥ 5 and ≤ 0.40. Only evidence-grade cohorts — same type and same market area — receive evidence exhibits; low-confidence flags showed no measured lift and are labeled and excluded.

Savings are cap-aware

Estimated savings = the reduction that actually reaches this year's taxable value, at the parcel's own composite rate (summed over its taxing entities). A market reduction below the taxable value doesn't save money this year but lowers the cap base — that is reported separately as future-year benefit. Both numbers appear on every exhibit.

The severe tier

Rather than discard extreme flags as probable data artifacts, we label them. A flag is severe when the subject's improvement value per square foot is at least three times (or at most one-third of) its cohort median, or when the subject is assessed more than 75% above its equity-indicated value — and it is not a special-use or mixed-use parcel (those remain quarantined for manual review). This is grounded in the Harris County calibration: on the protested basis, severe-tier parcels realized roughly 12–27% mean reductions versus ~9% for ordinary flags (Harris TY2024–25; 134 $/sf-outlier and 548 extreme-excess parcels — small samples, historical not predictive). Because an extreme flag can also reflect a genuine finish, quality, or construction-class difference from the cohort, every severe lead ships with a field-verification note directing the consultant to confirm true-peer status before filing.

Which values enter the median — the comparable-basis rules

The statute's median is only as good as the values behind it. Our rules, from the statute and the district's own practice:

• Appraised value, except capped parcels enter at market. §42.26(d) commands that where a §23.23 or §23.231 limitation applies, "the value of the property subject to the suit and the value of a comparable property … must be the market value." Applied symmetrically to subject and comps.
• No fully-exempt parcels, either side. Parcels wholly exempt (charitable, religious, housing-finance, community land trust families) carry roll values no one defends; they are removed from the pool entirely.
• Comp-side hygiene. Parcels with value-distorting partial exemptions (abatement, freeport, historic, pollution control), parcels in active district-court litigation (their values are provisional), and parcels whose improvement stack mixes types (dominant type < 70% of improvement value) are excluded as comparables.
• No self-comparison, no double counting. A subject never appears in its own comp set; undivided-interest parcels are consolidated before screening.

What it measured — Harris County, TY2024–2025

The identical screen, run on two prior years of noticed values, joined to HCAD's published ARB hearing outcomes. Formal hearings (conclusion codes FC/FN), evidence-grade cohorts, high confidence, protesters only:

All protesters in high-confidence cohorts, by representation status:

What a buyer receives

• Season Package — a season of prospecting, pre-armed. The complete ranked county list (1,479 flagged commercial parcels; 448 open prospects: 366 never-represented and 82 recently lapsed — each lapsed lead named with its prior firm and authorization end date). An owner-portfolio view groups parcels under common ownership — one signature can carry several protests — with a top-owners sheet leading the delivery. Evidence sheets for the 100 highest-ranked open prospects (never-represented and recently lapsed), prepared in your firm's name, each expandable to a formal-hearing packet; sheets for parcels another firm already represents come through Evidence Packets. The severe tier — the parcels whose excess historically produced the deepest realized cuts — labeled with its measured context. Includes the April 2027 notice-season re-screen (subject to county publication): the list re-ranks and the exhibits are rebuilt on the 2027 notice values for filing. Parcels in litigation, public and institutional owners, and assets over $25M are already removed.
• Evidence Packets — analyst hours, manufactured. For the book you already carry: a one-page equity exhibit per parcel — subject summary, adjusted comparable grid, median math, cap-aware savings, sources and statutes — in the form TCAD's informal channel consumes (documented evidence uploaded 48 hours ahead) — built on the 2026 certified roll as a prospecting exhibit, rebuilt on the 2027 notice values for filing (included), expandable to a six-page formal packet: owner-facing summary for the client conversation, comparable location map, per-comparable detail, methodology, verbatim statute, provenance. First five free on parcels you represent; then $1,500 per 50 at any scale your season needs. Download a watermarked sample →

Every deliverable ✻ re-verified against live county records at delivery · every figure cited to source and vintage · licensed under 4226tax.com/terms.

Validation

Every deliverable parcel is reconciled field-by-field to TCAD's certified export and a stratified sample is re-checked against the live county portal each season (our verification standard). Underneath that, the scoring is held to a from-scratch independent SQL recomputation (0 mismatches on the sampled sheets), deterministic hash-identical rebuilds, and mutation testing of the screen code (6,900 injected defects, zero survivors) — with the full method, counts, and taxonomy of every reconciled difference kept in the project's validation report.

Sources. Travis: TCAD appraisal export TY2026 (supp. 1 · run 2026-08-05 · layout v8.0.0.33), traviscad.org/publicinformation; TCAD agent-of-record 2023–2025, PIA delivery R010639-082426 (Aug 2026); Travis County Tax Office truth-in-taxation rate table (2025 adopted). Harris: HCAD PDATA real-property and ARB hearing files, TY2024–2026, download.hcad.org. Statute: Tax Code §41.43, §42.26, §23.231 (tcss.legis.texas.gov, retrieved 2026-08-22). Full calibration tables are delivered with every package.