Commercial property-tax appeals

Commercial real estate is routinely over-assessed, and the only fix is a deadline-bound, jurisdiction-specific appeal: pull comparable sales, build an income/cap-rate valuation, assemble an evidence packet, and argue it before an assessment board. Today boutique consultants and attorneys do this by hand on a contingency fee of 33–50% of the tax savings (Russell, Krafft & Gruber) — and they win 70–85% of the time when the evidence is good (Polter). That combination — provable ROI, recurring annual savings, and a value-capture model with zero budget objection — is the cleanest business case on the opportunity board. The agent owns the detection-to-filing case work; a licensed human still argues the hearing.

Vitals: market: $797B US property tax/yr (Census) · deadline-driven, per-jurisdiction · buyer: CRE owners / CFOs / consultants · model: contingency on savings · whitespace: ★★★

Market context — size, money flow, why CRE owners care

• The base: US state + local property tax revenue hit ~$797B in 2024, up 8.2% (Census via NAHB); property taxes are 70% of all local tax collections (Tax Foundation). Commercial/industrial property is a large slice of that base.
• The leakage: appeals succeed 70–85% with professional evidence (Polter) and 40–60% on average (NTU Foundation, reported) — i.e., over-assessment is systematic, so even a low single-digit error rate against an ~$797B base implies billions in recoverable tax.
• Why owners care beyond cash: an over-assessment doesn’t just cost annual tax — it depresses the asset. A $16K annual overpayment can knock $200K–$270K off property value at prevailing cap rates (Uthoff Graeber). NAIOP, the CRE-development association, openly urges owners to “contest excessive assessments” (NAIOP).
• The win is recurring: a single 50% assessment reduction produced $89,665 in tax savings in one documented case (Property Valuation Services), and a reduction lowers the base for subsequent years too.

The mess

• The clock is short, per-jurisdiction, and unforgiving. Appeals must usually be filed within 30 days of the assessment notice (some allow 45–60) (Madras), on dates that vary by jurisdiction (Florida: 25 days after the TRIM notice — CBIZ). Miss it and the owner eats the over-assessment for the whole year.
• Case-building is manual valuation work. Winning means pulling comparable sales and running an income/cap-rate analysis for income-producing property, then assembling a defensible evidence packet — the labor boutique consultants charge a third to half the savings to perform (rkg).
• It’s a two-track bureaucracy. Administrative “grievance” review and a board-of-review hearing first, judicial review (tax court) if that fails (NY Courts) — each with its own forms, evidence rules, and calendar.
• Coverage is the real gap. A consultant works the parcels an owner remembers to send. Nobody is monitoring every parcel in a portfolio every cycle for a fresh over-assessment — so winnable appeals silently lapse past the deadline.

Why now

The business model has always been clean — you bill from money you save, so there’s no budget fight and ROI is provable from day one. What was missing was the ability to do the consultant’s case work at software cost and software coverage. An agent can now ingest an assessment roll, pull and weigh comps, run an income/cap-rate valuation, draft the jurisdiction-specific appeal, and hit the deadline — across an entire portfolio, every cycle, not just the parcels someone flagged.

Why it stayed a boutique cottage industry: the work is deadline-bound and fragmented across thousands of assessing jurisdictions with their own forms, calendars, and evidence norms — exactly the per-jurisdiction arcana that resists generic software (Opportunities thesis). That fragmentation is the moat; an agent that encodes it scales where a consultant’s billable hours can’t.

The money

Signal	Figure	Basis
US property tax base	~$797B/yr (2024, +8.2%)	Census via NAHB (nahb) — VERIFIED
Local reliance on property tax	70% of local tax collections	Tax Foundation (tf) — VERIFIED
Commercial contingency fee	33–50% of (first-year) tax savings	Law-firm fee terms (rkg, pol) — VERIFIED
Win rate w/ professional evidence	70–85% (commercial); 40–60% avg	Polter (pol); NTU, reported (ntu) — VERIFIED / INFERRED
Single-appeal upside	$89,665 saved on one CRE reduction	Case study (pvs) — VERIFIED
Asset-value stakes	$16K/yr tax → $200–270K off value	Cap-rate math (uth) — VERIFIED

The cleanest model on the board

You’re paid a cut of tax you actually save, so the buyer is self-funded and ROI needs no projection — and the saving recurs because a lowered assessment resets the base for future years. This is the genuine contingency the probate opportunity lacks.

How it works today

A consultant works one parcel at a time, on the owner’s prompt, against a hard deadline. The valuation is built by hand; the win recurs but the coverage depends on someone remembering to file.

Mermaid source

flowchart LR
  classDef human fill:#fdecec,stroke:#e0564f,stroke-width:1.5px,color:#0f172a;
  classDef pro fill:#eef0fe,stroke:#6366f1,stroke-width:1.5px,color:#0f172a;
  classDef gov fill:#e8f1fd,stroke:#2563eb,stroke-width:1.5px,color:#0f172a;
  classDef ext fill:#fff7ed,stroke:#d97706,stroke-width:1.5px,color:#0f172a;

  Assr("County assessor<br/>sets assessed value"):::gov
  Notice("Assessment notice<br/>~30–45 day appeal window"):::gov
  Owner("CRE owner / CFO<br/>often misses it or eats it"):::human
  Cons("Boutique consultant / attorney<br/>contingency: 33–50% of savings"):::pro

  subgraph Work["Manual case-building, by hand"]
    direction TB
    Comps("Pull comparable sales"):::ext
    Income("Income / cap-rate analysis"):::ext
    Pkt("Assemble evidence packet"):::pro
  end

  Board("Board of review /<br/>assessment appeals board<br/>hearing"):::gov
  Court("Tax court<br/>judicial review"):::gov
  Save("Reduced assessment<br/>→ recurring annual tax savings"):::human

  Assr --> Notice --> Owner --> Cons
  Cons --> Comps
  Cons --> Income
  Comps --> Pkt
  Income --> Pkt
  Pkt --> Board --> Save
  Board -. "escalate" .-> Court

Where an agent fits

Flip coverage from reactive to continuous. The agent watches every parcel across every jurisdiction, flags fresh over-assessments, and — for the ones worth pursuing — builds the comps-plus-income valuation, generates the jurisdiction-specific appeal packet, files before the deadline, and tracks the hearing calendar. The owner (or a licensed agent) still appears at the board hearing where representation is required; everything upstream is the agent’s. Because it’s billed on tax saved, the audit-ready evidence packet is both the work product and the meter.

This is the playbook in miniature: an agent that acts on the deadline rather than waiting to be asked; per-jurisdiction rules — deadlines, forms, evidence norms — encoded as a domain layer instead of hand-held per county; and pulling assessor rolls and comps from systems with no API.

Mermaid source

flowchart LR
  classDef human fill:#fdecec,stroke:#e0564f,stroke-width:1.5px,color:#0f172a;
  classDef agent fill:#eafbf1,stroke:#16a34a,stroke-width:1.5px,color:#0f172a;
  classDef ext fill:#fff7ed,stroke:#d97706,stroke-width:1.5px,color:#0f172a;
  classDef store fill:#eef0fe,stroke:#6366f1,stroke-width:1.5px,color:#0f172a;

  Owner("CRE owner / portfolio"):::human

  Agent("Property-tax appeal agent<br/>watches every parcel + jurisdiction<br/>acts on the deadline, not on request"):::agent

  subgraph Owns["What the agent owns"]
    direction TB
    Mon("Monitor assessments<br/>flag over-assessment"):::agent
    Val("Build comps + income /<br/>cap-rate valuation"):::agent
    Pkt("Generate jurisdiction-specific<br/>appeal packet · meet deadline"):::agent
    Track("File + track hearing"):::agent
  end

  Data("Assessor rolls · sales comps · rent / cap data<br/>per-jurisdiction, mostly no API"):::ext
  Gate{{"Licensed rep at hearing<br/>(human gate)"}}:::human
  Bill[("Billed on tax saved · contingency<br/>audit-ready evidence packet")]:::store

  Owner --> Agent --> Owns
  Owns -. "reads" .-> Data
  Owns --> Gate
  Agent -. "evidence + meter" .-> Bill

Whitespace & incumbents

Residential is no longer open; commercial still is. The field sorts cleanly:

• AI-native, but residential-first — Ownwell is the funded entrant: it “manage[s] the end-to-end process of property tax protests and appeals… from paperwork to negotiations and appeal hearings” (Ownwell) and raised $50M in Feb 2026, reporting “86% success and $774 average savings per customer” (HousingWire). But that $774 figure is a homeowner number — Ownwell appeals taxes “on behalf of homeowners” (Crunchbase). The income-approach valuation that wins commercial appeals is a different, harder product.
• Commercial incumbents are consultancies, not software — Ryan LLC and peers (Altus, Paradigm) dominate commercial property tax, pairing contingency consulting with “web-based property tax software to track your commercial property tax history by property and by parcel” (Ryan). The software tracks; the humans still do the valuation and the hearings.

The real risk is a residential AI moving up-market

Commercial CRE appeals are genuinely under-tooled by AI today (★★★) — but the obvious threat isn’t the boutiques, it’s a well-capitalized residential player like Ownwell extending into the income-approach work. The defensible wedge is owning that commercial valuation + per-jurisdiction filing depth before they do.

Hard problems

Problem	Why it’s hard here	Signal	Likely approach (speculative)
Commercial valuation, not comps lookup	Income-producing CRE is valued by income/cap-rate, not a sales-comp pull; the agent must reason about NOI, vacancy, and cap rates to a defensible standard	cap-rate math underpins the asset-value case (uth); win rate hinges on “professional evidence” (pol)	Probably a valuation model (income + sales + cost approaches) that outputs an evidence packet a licensed appraiser/agent will sign, not a single number
Per-jurisdiction deadlines & forms	Thousands of assessing jurisdictions, each with its own window, forms, and evidence rules; one missed date forfeits the year	30/45/60-day windows (mad); FL “25 days after TRIM” (cbiz)	Likely a per-jurisdiction rules/calendar layer (deadlines + form templates as a DSL) — see encoding domain rules
Sourcing assessor + market data	Assessor rolls, sales comps, and rent/cap data are fragmented and mostly API-less, varying by county	Ryan sells “by property and by parcel” tracking software because the data is messy (ryan)	Probably scraping/ingest of assessor portals + comps feeds, normalized per jurisdiction — see integrating systems without APIs
The hearing still needs a human	Many jurisdictions require a licensed agent/attorney to represent at the board; the agent can’t fully close the loop alone	Two-track admin→judicial process (nyc); consultants/attorneys hold the relationships	Likely agent-does-everything-to-the-hearing, with a licensed rep (in-house or partner) gating the appearance and signature

Sources

• Census via NAHB · Tax Foundation — property-tax market size and local reliance
• Russell, Krafft & Gruber · Polter · Peraica retainer — commercial contingency fees and win rates
• NTU Foundation, reported — national appeal success rate
• Property Valuation Services · Uthoff Graeber · NAIOP — savings cases and asset-value stakes
• Madras · CBIZ · NY Courts — deadlines and the two-track process
• Ownwell · HousingWire · Crunchbase · Ryan LLC — incumbents and the AI-native entrant

Reconstructed from public sources; claims are tier-labeled (VERIFIED / INFERRED / SPECULATIVE) — see how to read the tiers. Supporting quotes live in this repo’s evidence map (evidence/opp-property-tax-appeals-evidence-map.md).