CompForge

1. One-liner

Turns an insurer’s lowball CCC total-loss valuation into a deadline-aware, comp-backed appraisal-clause rebuttal in minutes.

2. Trend signal — why now?

A regulatory door just opened and the people who walk through it have no software.

As of January 1, 2026, Texas SB 458 (Insurance Code Ch. 1813) requires every personal auto and residential property policy issued or renewed in Texas to carry a binding appraisal clause. Crucially, either party can demand appraisal unilaterally — no mutual agreement, no need to prove an impasse — with hard deadlines: 20 days to name an appraiser, 120 days for appraisers to agree, 240 days to a final award. The legal commentary is blunt: this “will probably do more to increase appraisal volume in Texas than any other element of the rule.” Washington passed its own right-to-appraisal law (effective July 2025, with the OIC now standing up an umpire registry), and Illinois HB 4160 cleared the House Insurance Committee in March 2026.

Meanwhile the thing being disputed — the insurer’s CCC ONE market valuation — is widely documented as gameable: cherry-picked lower-trim comps, geographically distant listings, one-directional condition deductions, and stale per-mile depreciation. A cottage industry of independent appraisers already charges $275–$300 per rebuttal report (plus another ~$275 for negotiation rep) to fight these by hand. The work is “complex and time-consuming” — pull the multi-page CCC report, audit each of the six comps for whether it’s even currently listed, hand-find 3–5 genuine local matches, and write the narrative. That’s 1.5–2.5 hours per file.

So: a new law floods the dispute side with volume on a stopwatch, the disputed artifact has known, repeatable flaws, and the only tooling that exists (CCC, Mitchell, Audatex) is sold to the insurers. The dispute side runs on spreadsheets and browser tabs.

Provenance:

• Signal 1 (demand): Independent appraisers charge $275–300/report to manually rebut CCC ONE valuations; documented CCC flaws (cherry-picked comps, distant markets, unfair condition/mileage adjustments) — https://snapclaim.com/ccc-one-market-valuation-report-flaws/ — 2026-06-27
• Signal 2 (economic): TX SB 458 makes appraisal unilateral with hard deadlines eff. Jan 1 2026; WA RTA law live; IL HB 4160 advancing — TDI itself expects a step-change in appraisal volume — https://www.gavnat.com/blog/texas-appraisal-law-updates-for-2026-what-sb-458-changes-for-homeowners-insurers-and-public-adjusters/ — 2026-06-27
• Signal 3 (feasibility): MarketCheck Cars API exposes 540M retail listings / 5B+ historical, self-serve, real-time — enough to auto-source verifiable local comps; LLM PDF extraction makes parsing the CCC report turnkey — https://www.marketcheck.com/apis/cars/ — 2026-06-27 Category: Regulatory arbitrage

3. The opportunity

The total-loss dispute is a structured argument with a known opponent: the CCC ONE report. Every rebuttal does the same four things — challenge comp selection, challenge geographic relevance, challenge condition adjustments, challenge mileage math — then assemble local listings that prove a higher number, then narrate it for an appraiser or umpire who has 240 days to rule.

The incumbents (CCC Intelligent Solutions, Mitchell, Solera/Audatex) build valuation engines for insurers. Their entire economic incentive is to produce the number being disputed. They will never ship a tool that attacks their own output. The dispute side is served only by done-for-you service shops (SnapClaim, Fair Auto Appraisals, AutoClaim Consultants) — which means an appraiser doing 30 files a month is either one of those shops doing it by hand, or a solo competing against them with no leverage.

CompForge sells the leverage, not the service. Upload the CCC PDF, get back a near-finished, citation-backed rebuttal packet plus a deadline tracker tied to the appraisal-clause timeline. The appraiser 4בs throughput and keeps the $275–300 client fee. That’s a tool a professional pays for this month because the law just multiplied their backlog.

4. Target market

• Primary customer: US independent auto-damage appraisers and small public-adjuster / claims-consulting shops (1–5 people) who do total-loss and diminished-value disputes. BLS counts ~9,200 auto-damage appraisers (2024); the dispute-focused subset is the wedge, concentrated first in Texas and Washington where the 2026 mandate just hit.
• Why they buy: In their words, disputing a CCC report is “a complex and time-consuming process” and they want someone to “handle the legwork.” Each file is 1.5–2.5 hrs of comp-hunting and write-up; the new unilateral-demand rule means more files arriving with a 240-day clock. They already monetize this at $275–300/report — software that triples capacity is pure margin.
• Rough TAM reasoning: Conservatively 2,000–4,000 active dispute-side appraisers/adjuster shops nationally today, expanding as TX/WA/IL normalize unilateral appraisal. At $149/mo that’s a $3.5M–$7M ARR ceiling on the core tool alone — squarely in the bootstrap zone — before per-report or property-line expansion.
• Why now for them: TX SB 458’s unilateral-demand + hard-deadline structure (eff. Jan 1 2026) turns appraisal from a rare escalation into a routine, time-boxed workflow. Volume up, clock running, no purpose-built tool.

5. Product sketch (MVP)

• Drop the CCC ONE / Mitchell PDF in — it extracts the subject vehicle (VIN, trim, options, mileage, condition) and all six insurer comparables into a structured grid.
• Comp integrity audit — for each insurer comp, checks whether it’s currently and verifiably listed, flags trim/option mismatches, salvage/accident history, and distance from the loss ZIP, with a one-line “why this comp is weak” for each.
• Live local comp pull — queries a listings API for genuine current matches by trim/options/mileage within a defensible local radius; ranks the strongest 3–5.
• Condition & mileage rebuttal — recomputes the adjustments CCC applied and surfaces the ones that are one-directional or use generic per-mile rates.
• Auto-drafted rebuttal narrative — generates the written argument (editable) citing each flagged flaw and each supporting listing, formatted for submission to the opposing appraiser or umpire.
• Appraisal-clause deadline tracker — per-file timeline (20/120/240-day milestones for TX; configurable per state) with reminders so a missed statutory deadline never costs an award.
• Branded export — clean PDF packet on the appraiser’s letterhead, ready to send.

6. AI angle — what’s load-bearing

Remove the AI and there is no product — it collapses back to the 2-hour manual file. Three places it does real work:

1. Document understanding: parsing a messy multi-page CCC/Mitchell PDF into a clean, structured subject-vs-comps grid is exactly the vision/LLM extraction task that became reliable and cheap in the last year.
2. Comp matching & weakness scoring: judging whether an insurer comp is a true match (same trim, options, real local listing) vs. a cherry-pick is fuzzy, language-heavy reasoning over listing text — not a SQL join. The model scores each comp and explains why it’s weak.
3. Argument drafting: turning the flagged flaws + supporting listings into a persuasive, professional rebuttal narrative is the highest-leverage minutes saved. This is generation grounded in the structured evidence the tool already assembled — citations, not hallucinations.

7. Localization angle (if any)

N/A — this is a US-only play by construction. The entire opportunity is the localization: it’s pinned to US state appraisal-clause law (TX SB 458, WA RTA, IL HB 4160) and the US-specific CCC/Mitchell valuation ecosystem. The “localization” wedge here is state-by-state regulatory tracking — encoding each state’s appraisal deadlines and qualification rules into the deadline engine — which is a moat, not a translation chore.

8. Business model — path to $1M–$5M ARR

• Pricing: $149/mo per appraiser seat (Starter, ~20 reports/mo) and $349/mo (Pro, unlimited + diminished-value module). Optional metered overage / pay-as-you-go at $19 per generated packet for low-volume solos. Listings-API cost (a few dollars per report) is baked in.
• ACV: ~$2,200/yr blended (mix of Starter, Pro, and metered).
• Math to $1M ARR: ~455 paying seats × ~$183/mo avg × 12 ≈ $1.0M. That’s a single-digit percent of the active dispute-side population — achievable.
• Math to $5M ARR: ~1,900 seats, or fewer seats plus the property line. SB 458 covers residential property too; the same engine (challenge the carrier’s estimate, assemble counter-evidence, track deadlines) extends to home/roof total-loss disputes where public adjusters charge far more — that’s the ACV expander.
• Expansion path: seats → diminished-value module → property/residential module → state coverage upsell (each new RTA state = new addressable seats) → per-packet credits for occasional users who later convert to seats.

9. Go-to-market wedge — first 100 customers

• Hit the state license rosters. Texas and several states publish lists of licensed/registered auto damage appraisers and public adjusters. Scrape TX + WA rosters (the two live-mandate states), filter to dispute-side shops, send a personalized 2-minute Loom showing their actual market getting a CCC report turned into a rebuttal. Expect 3–5% reply on a well-targeted, timely list.
• Ride the rule change. Comment on / publish in the venues where this audience already gathers around SB 458 — Repairer Driven News, Autobody News, the TDI rule-hearing crowd. A “here’s the new 240-day clock and a tool that hits every deadline” piece lands when the pain is fresh and dated June 2026.
• Convert the service shops’ overflow. SnapClaim, Fair Auto Appraisals, AutoClaim Consultants and dozens of regional shops are the buyers — they do this by hand at scale. Cold-outreach the ~200 visible dispute-service shops directly with a throughput-and-margin pitch (3× files per appraiser, same fee).
• Diminished-value adjacency. The DV appraiser community (BOCAA-certified appraisers, DV-focused shops) does near-identical comp work and is densely networked — a referral/affiliate motion through that community seeds the first cohort.

10. Build complexity — justification

Medium. The hard parts are off-the-shelf: PDF/vision extraction via current LLM APIs, comps via the MarketCheck Cars API (self-serve, real-time, 540M+ listings), narrative via standard generation. The genuinely custom work is the comp-weakness scoring logic and the state-by-state deadline/qualification engine — both are bounded rules-plus-model work, not research. A 2-person team ships a credible TX-only v1 in ~12–16 weeks; the moat-building (more states, property line, tuning against real umpire outcomes) is the ongoing work after launch.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	A drafting/evidence tool for licensed appraisers; doesn’t itself render legal opinions or require a license. Must avoid UPL — it assists the professional, not the consumer-as-litigant.
Ethical — no harm / dark patterns	✅	Counterweight to documented insurer undervaluation; pro-consumer outcome via the professional.
Market exists (evidence above)	✅	$275–300 paid per manual report today; new law multiplies volume.
1–5 person team can build this	✅	2 people, ~12–16 wks to TX v1 on off-the-shelf APIs.
Launchable with <$50K / ₹40L	✅	Main spend is API usage + a designer; well under $50K.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Hair-on-fire for the appraiser the moment a 240-day clock starts; they already pay to solve it. Not daily for any one shop, hence not 18+.
Demand evidence	15	12/15	Hard signals: paid services at $275–300, documented CCC flaws, named law, growing dispute volume. Knock from not having direct appraiser-side quotes (Reddit/forums inaccessible this run).
Build feasibility	15	11/15	Off-the-shelf APIs do the heavy lifting; comp-scoring + multi-state deadline engine is real but bounded work.
Distribution clarity	15	11/15	Named, scrapable license rosters + visible service shops + DV community. Conversion math is estimated, not proven.
Revenue mechanics	15	12/15	Pricing anchored to an existing $275–300 per-report economy; ARR math needs only single-digit % penetration. Some reliance on property-line expansion for the $5M case.
Time to first revenue	10	8/10	Self-serve SaaS for pros with acute, dated pain → paid pilots within weeks of a TX-targeted launch.
Defensibility	10	6/10	Soft moat: state-deadline/regulatory knowledge + workflow lock-in + outcome data accumulating against umpire rulings. Copyable, but the multi-state rule engine and head start are real.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required — needs solid PDF/LLM engineering plus a founder or advisor who actually knows the appraisal-clause workflow and can encode state rules correctly.

Key assumptions to validate (3–5)

1. Assumption: Dispute-side appraisers will pay ~$149/mo for software vs. continuing by hand. How to test: 25 cold calls/Looms to TX + WA roster appraisers; pre-sell 5 annual pilots before building beyond a demo.
2. Assumption: CCC/Mitchell PDFs extract cleanly enough for a reliable comp grid across makes/models. How to test: Run 50 real (anonymized) CCC reports through an extraction prototype; measure field accuracy, target >95% on subject + comps.
3. Assumption: MarketCheck (or equivalent) listing coverage is dense enough to find defensible local comps for typical loss vehicles. How to test: For 50 sample losses, check that the API returns ≥3 genuine current local matches; fall back to multi-source if coverage is thin.
4. Assumption: Auto-drafted rebuttals are persuasive enough that appraisers ship them with light edits. How to test: Blind-rate 20 generated narratives with 3 working appraisers against their own write-ups.

Risk flags

1. Platform/data dependency: Reliance on a single listings API (MarketCheck) and on CCC’s PDF format staying parseable. Mitigation: multi-source listings, and CCC format changes are slow and detectable.
2. Regulatory/UPL risk: Must stay a tool for licensed professionals; selling rebuttal-generation directly to consumers risks unauthorized-practice and licensing exposure in some states. Mitigation: B2B-only positioning, professional-in-the-loop.
3. Market timing / scope creep: TX/WA are live but most states aren’t yet; if RTA momentum stalls (IL bill dies), the addressable map grows slower. Mitigation: the manual pain already exists nationwide via the standard policy appraisal clause — the mandate accelerates, it isn’t the only demand.
4. Incumbent response: CCC/Mitchell could add a “fairness” feature, but it cannibalizes their insurer business — low probability, which is the structural opening.

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + appraisal-clause domain advisor (TX/WA)
Time to revenue:        6–10 weeks from a TX-targeted launch
Capital to launch:      $15–30K (mostly API usage + design)
Top 3 assumptions to validate first:
  1. Appraisers pay ~$149/mo vs. doing it by hand — pre-sell 5 annual pilots from the TX/WA roster
  2. CCC/Mitchell PDFs extract to a reliable comp grid — 50-report accuracy test, target >95%
  3. Listings API returns ≥3 defensible local comps per loss — 50-loss coverage test
Kill criteria:
  - Abandon if <2 of 25 targeted TX/WA appraisers will pre-pay a pilot
  - Abandon if PDF extraction accuracy on real CCC reports stays below ~90% after tuning
  - Abandon if CCC or Mitchell ships a free policyholder-side rebuttal feature before v1

15. Next step — 1-week validation sprint

• Day 1–2: Scrape the Texas + Washington licensed-appraiser / public-adjuster rosters; build a 150-name dispute-side shortlist. Collect 30 real (anonymized) CCC ONE reports from public dispute write-ups and any friendly appraiser.
• Day 3–4: Build a throwaway extraction + comp-pull prototype (CCC PDF in → subject/comp grid + 3 MarketCheck local comps + a drafted rebuttal paragraph out). Measure extraction accuracy on the 30 reports.
• Day 5: Send 25 personalized Looms showing a real rebuttal generated for the appraiser’s own market; ask for a $149/mo annual pilot. Go if ≥3 pre-pay or give a verbal LOI; no-go if <2 and extraction accuracy is below 90%.

Falsifiable outcome: pre-paid pilots and a measured extraction-accuracy number — not vibes.