LightCall

1. One-liner

Tells a used-car manager which sale light and announcements a car needs before it hits the block.

2. Trend signal — why now?

On June 1, 2026, the National Auto Auction Association’s updated Arbitration Guidelines took effect, and they moved the risk decisively onto the seller:

• Generic disclaimers are dead. “No Arb for [defect]” announcements are now prohibited and may themselves be grounds for arbitration. Manheim’s own guidance to sellers: “Statements like ‘No Arb for Oil Leaks’ will no longer be sufficient.”
• Warning lights no longer count as disclosure. Announcing only a warning light or diagnostic trouble code “no longer relieves a seller of responsibility for the underlying condition.” You have to name the actual defect.
• ADAS became arbitrable. Adaptive cruise, lane departure warning, blind spot monitoring are now formally eligible for arbitration claims — a category most sellers have never announced on in their lives.
• A graduated threshold on high-value units. At $50,000+, each individual defect must hit 2% of purchase price; under $50,000 the $800 threshold stands.

This lands on a workflow where the seller was already holding the bag. ADESA’s guidelines state it plainly: “Seller is responsible for the accuracy and completeness of all representations, announcements, and disclosures regarding its vehicles, regardless of whether Seller has relied on third-party resources.” Translation: the auction’s own condition report being wrong is not your defense.

Meanwhile the defect surface is exploding. ADAS calibrations appeared on 34.7% of repair estimates in 2025, up from 12.1% in 2022, averaging $688 when present — comfortably over the $800 threshold once a second system is involved, and squarely inside the newly-arbitrable category.

Claim volume is not trivial: roughly 3–5% of B2B auction transactions result in arbitration claims, against a US vehicle auction market that moved 14.26M units in 2024. Losing costs real money — ADESA charges $195 per claim, and per its guidelines “the unsuccessful party in the arbitration will be responsible for payment of all fees (including but not limited to the inspection, diagnostic evaluation and transportation).” Copart’s policy reserves the right to charge the seller the arbitration fee, transportation fee, and repair estimate.

Provenance:

• Signal 1 (demand): NAAA Arbitration Guidelines effective June 1, 2026 ban generic “No Arb” announcements, void warning-light-only disclosure, and make ADAS arbitrable — shifting disclosure burden onto sellers — https://niada.com/dashboard/naaa-sharpens-arbitration-rules-on-announcements-high-value-vehicles-and-adas/ — 2026-08-27
• Signal 2 (feasibility): Smartphone AI vehicle-damage detection is commodity — Monk AI (acquired by ACV Auctions), Ravin AI, Click-Ins, PAVE, Inspektlabs all sell photo-based condition detection via API; Cox/vAuto shipped a UVeye integration for appraisal — https://inspektlabs.com/blog/top-10-ai-powered-car-damage-inspection-solutions-2/ — 2026-08-27
• Signal 3 (economic): ADAS calibrations on 34.7% of 2025 estimates vs 12.1% in 2022 at $688 average; 3–5% of B2B auction transactions arbitrate; ADESA charges $195/claim and losers pay inspection + transport — https://www.autoauctionatlas.com/learn/guides/arbitration-claims/ — 2026-08-27 Category: Workflow automation (a high-frequency judgment call made under time pressure in a $1B+ industry) + Regulatory arbitrage (a June 2026 rule change that redefined what a seller must say)

3. The opportunity

Every vehicle a dealer wholesales requires a decision that nobody has tooling for: what light do I run it under, and what exactly do I announce?

Get it wrong in the conservative direction — red light everything — and you leave money on the table, because a red-light car sells for less. Get it wrong in the aggressive direction — green light, minimal announcements — and you eat an arbitration claim: the unit comes back, you pay the claim fee, the inspection, the transport, and your seller rating degrades, which affects your fee structure on every future unit.

Until June 2026 there was an escape hatch: announce vaguely, slap “No Arb” on the known problem, move on. That hatch is now welded shut, and a whole new category (ADAS) got added to the arbitrable list on the same day.

Who solves this today? Nobody, on the sell side. The AI-vision field is real and mature — Monk, Ravin, Click-Ins, PAVE, Inspektlabs, UVeye — but every one of them sells detection: here are the dents, here is the damage map. They sell to auctions, insurers, rental fleets, and to dealers for appraisal and sourcing (that’s exactly what the Cox/vAuto–UVeye integration is for — buying better, not selling safer).

The one company sitting near arbitration, ReconCash, is explicitly buy-side: a “nationwide arbitration inspection service” that helps dealerships file and recover on cars they bought. Nobody is on the other side of the table helping the seller not get hit in the first place.

That’s the gap: detection is solved, the announcement decision is not. Going from “there is a code stored for the blind-spot module” to “under the June 2026 guidelines you must announce the blind-spot monitor as inoperative, and the DTC alone will not protect you” is a rules-mapping problem, not a computer-vision problem — and it’s the part the seller actually gets judged on.

4. Target market

• Primary customer: Used-car manager or wholesale manager at an independent used-car dealership or a small franchise store in the US, wholesaling 15–80 units a month. The buyer is the person who personally signs off on the light and the announcement sheet.
• Why they buy: Because an arbitration comes back to them. The forum voice is consistent about how the process feels from the losing end — from DealerRefresh: “They deny it because there was no picture of it in the CR. WTH?” and “You will not get it no matter how many times you ask or how many claims you open.” The lesson dealers draw is defensive and expensive: “Next time just factor in that the cable might not be there and adjust your bid accordingly.” On the sell side, the equivalent of that lesson is red-lighting everything and eating the price haircut.
• Rough TAM reasoning: ~52,000 active independent used-car dealers in the US, plus 16,990 franchised dealerships that all wholesale their aged and off-profile inventory. Wholesale is the standard exit for aged units — one Midwest group posts to online auction at 45 days in inventory. Even a narrow beachhead of the 13,000+ NIADA member dealers at $250/mo is a $39M ceiling; I only need a slice.
• Why now for them: The rules changed under them on June 1, 2026, and most of them found out from a trade newsletter, not from training. The announcement habits that worked for a decade now create liability.

5. Product sketch (MVP)

• Walk-around capture — the lot porter shoots a guided photo sequence and an OBD scan on a phone; no new hardware.
• Announcement sheet — for each unit, the specific announcements the June 2026 guidelines require, in NAAA-compliant language, ready to paste into the consignment form.
• Light recommendation with a price delta — green / yellow / red, with the arbitration-risk reasoning and an estimate of what the light choice costs or saves you on this unit.
• Threshold calculator — applies the $800 rule under $50K and the 2% rule at $50K+, per defect, so you know which findings actually clear the bar and must be announced.
• ADAS checklist — flags which driver-assist systems the vehicle is equipped with and whether each is verified working, since these became arbitrable in June 2026.
• Generic-disclaimer catcher — blocks “No Arb for X” phrasing and rewrites it into a specific, compliant disclosure.
• Defense file — timestamped photos, scan results, and the exact announcement text, retained per unit. When a claim lands inside the 10-day window, you have the evidence in one click.
• Rule-change feed — when NAAA or a specific auction updates policy, affected announcement templates update and you get told what changed.

6. AI angle — what’s load-bearing

Two places, both load-bearing:

Vision + scan interpretation. Turning a phone walk-around and an OBD dump into a structured defect list with severity and repair-cost estimates. This is the part that’s cheap now and wasn’t five years ago — and I’d buy it, not build it, at least initially (Inspektlabs, Ravin, and others sell this as an API).

Rules mapping — this is the actual product. Taking a defect list and answering: does this clear the threshold for this sale price, is it in an arbitrable category under the current guidelines, does the auction I’m consigning to have a stricter house rule, and what is the exact sentence I should announce? That’s reasoning over a corpus of NAAA guidelines plus per-auction policy documents that change on their own schedules, applied to a specific unit.

Remove the AI and this is a PDF checklist — which is precisely what exists today and precisely why dealers get it wrong. The value is that it runs per-unit, in 90 seconds, on the 40 cars going out this week, and it stays current when the rules move.

7. Localization angle (if any)

N/A — this is a US-first play. The wedge is a US institution: NAAA’s guidelines and the sale-light convention. The mechanic doesn’t transplant cleanly — BCA in the UK runs a 48-hour/500-mile window and CarOnSale in the EU runs a non-binding complaints process at €99 per validated complaint, both structurally different regimes. Those are later markets with a rebuilt rules engine, not a translation.

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

• Pricing: $199/mo for up to 25 units, $349/mo up to 60, $599/mo unlimited single rooftop. Overage $6/unit. Priced against the thing it replaces — a single lost arbitration on a mid-value unit runs $195 in claim fees plus inspection, transport, and the repair delta, so the tool pays for itself on roughly one avoided claim a quarter.
• ACV: ~$3,600/year blended.
• Rough math to $1M ARR: 280 rooftops × $299/mo × 12 = $1.0M. Against 52,000 independent dealers, that’s 0.5% penetration.
• Rough math to $5M ARR: ~1,200 rooftops, plus group-level deals. Realistically this needs the dealer-group channel — a 20-rooftop group at $500/rooftop/mo is $120K ARR from one sale — and an expansion into the buy-side mirror product (screening cars you’re bidding on against the same rule set), which doubles seats without new distribution.
• Expansion path: Per-rooftop seats within groups → buy-side screening module → arbitration defense-file retrieval as a paid tier → white-label to independent auctions who want their consignors to announce better (fewer claims is good for the auction too).

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

1. The rule change is the cold-email hook, and it has a shelf life. NIADA has 13,000+ member dealers and just published the “NAAA Sharpens Arbitration Rules” briefing. Build a list of independent dealers from state IADA member directories and dealer-license public records, and send one email: “Three things you could announce before June 1 that will now get you arbitrated.” Attach a free per-unit audit of their last 10 wholesale units. This converts because it’s news they half-heard and haven’t operationalized.
2. State IADA associations as a channel, not a logo. There are dozens of state-level independent dealer associations running monthly meetings and compliance seminars. They are actively looking for June-2026-guidelines content to program. Offer to give the talk free; the talk is the demo. Target 8 associations in the first two quarters.
3. Auction-lane intercept. Independent auctions (not Manheim/ADESA — the ~300 smaller regional houses) have a direct interest in their consignors announcing correctly, because every claim costs them staff time and buyer goodwill. Pitch 20 regional auctions on offering LightCall to their consignor base, revenue-shared. One auction with 200 active consignors is a channel, not a customer.
4. Post-claim retargeting. Dealers who just lost an arbitration are the highest-intent buyers alive. DealerRefresh, dealer Facebook groups, and NIADA forums have a steady flow of these posts. Answer the thread with the specific guideline that would have protected them, then DM.
5. The 20-group land-and-expand. Once ~30 single rooftops are live with claim-rate data, take that data to used-vehicle directors at 20-plus-rooftop groups. “Your stores arbitrate at X%; our cohort runs at Y%” is the only pitch that works at that level, and it requires having shipped first.

10. Build complexity — justification

Medium. The vision and OBD interpretation layer is bought, not built — several vendors sell photo-based damage detection as an API, which removes the hardest engineering. The custom work is the rules engine: encoding NAAA guidelines plus the house policies of the major auctions into per-unit announcement logic, and keeping it current. That’s a domain-modeling problem with a maintenance tail, not a research problem.

Realistic v1 for two people: 12–16 weeks. Capture flow and defect extraction in weeks 1–6, rules engine and announcement generation weeks 6–12, defense-file retention and auction-specific profiles weeks 12–16. The unglamorous risk is that the rules corpus needs a human who genuinely knows wholesale — budget for a part-time domain advisor from day one.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helps sellers comply with private association guidelines. No licensing required.
Ethical — no harm / dark patterns	✅	The product pushes toward more disclosure, not less. Aligned with buyers and auctions.
Market exists (evidence above)	✅	14.26M auction units, 3–5% claim rate, per-claim fees, published rule change.
1–5 person team can build this	✅	Two people plus a domain advisor; vision layer is a bought API.
Launchable with <$50K / ₹40L	✅	Main costs are per-inspection API calls and the domain advisor.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	15/20	Real money per event and the seller explicitly owns the liability “regardless of whether Seller has relied on third-party resources.” But it’s a 3–5% event, not daily pain — the median dealer feels this monthly, not hourly, and has absorbed it as a cost of doing business. That ceiling is honest.
Demand evidence	15	12/15	Strong published signals: dated rule change, claim-rate and fee data, ADAS trend from 12.1%→34.7%. Docked because the customer voice I found is buy-side dealers complaining about arbitration; I have no verbatim quote from a seller asking for announcement help. That’s the gap to close first.
Build feasibility	15	11/15	Vision is off-the-shelf. Rules engine is real work with a maintenance tail. 12–16 weeks for a pair, not 6.
Distribution clarity	15	12/15	Named lists (state IADA directories, dealer-license records), named associations, named intercept points. The regional-auction channel is genuinely promising. Docked because dealer cold email is a well-farmed channel with mediocre response rates.
Revenue mechanics	15	11/15	$199–599/mo is well inside dealer software norms and the ROI story is arithmetic. Docked because 280 rooftops for $1M means real sales throughput, and per-inspection API cost eats into margin at the unlimited tier.
Time to first revenue	10	8/10	Pre-sellable on the rule change before the product is finished; a manual-service version can bill in week 3.
Defensibility	10	5/10	Execution moat plus an accumulating rules corpus and per-dealer claim history. But Cox or ACV could bolt this onto existing dealer software, and ACV already owns Monk. A 12-month head start with real claim-rate data is the realistic defense, not a moat.
Total	100	74/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

The rules engine is worthless if it’s wrong, and it will be wrong unless someone on the team has actually sat in the used-car manager’s chair. A technical founder without a wholesale partner should not build this.

Key assumptions to validate (3–5)

1. Assumption: Sellers experience arbitration frequently enough to feel it — a dealer wholesaling 40 units/month gets hit often enough to want prevention. How to test: Interview 25 used-car managers; ask for their actual arbitration count over the last 12 months and what each one cost. If the median is under 3 a year, the pain is too rare to price at $299/mo.
2. Assumption: The announcement decision is genuinely uncertain, not obvious to an experienced manager. How to test: Blind-test 15 managers on 10 real units with defect lists; compare their light-and-announce calls to each other and to a guidelines-correct answer. Meaningful disagreement between experienced people proves the product; unanimous agreement kills it.
3. Assumption: Dealers will change the pre-sale workflow to add a 90-second step. How to test: Run the manual-service version — they text photos, we return the announcement sheet — for 10 dealers over 4 weeks and measure whether usage persists past week 2 without prompting.
4. Assumption: Regional auctions will co-sell to their consignors. How to test: Pitch 10 independent auctions; look for 2+ willing to email their consignor list.
5. Assumption: Photo + OBD capture is sufficient to catch the defects that actually get arbitrated. How to test: Take 30 units that were arbitrated in the last year and check retrospectively whether the capture flow would have surfaced the claimed defect. Structural damage and title issues — ~35% of claims combined — may be outside what a walk-around can see.

Risk flags

1. Incumbent absorption: ACV Auctions owns Monk AI and already runs the inspection layer; Cox owns vAuto and Manheim and just integrated UVeye. Either could ship a seller-side announcement assistant as a feature. The counter is that both are structurally conflicted — they operate the auctions that arbitrate, so tooling that helps sellers dodge claims is awkward for them. That conflict is the opening, and it’s real but not permanent.
2. Frequency risk (the big one): At a 3–5% claim rate, a dealer doing 40 units/month sees roughly 1–2 claims a month. That’s enough to price against, but it’s not hair-on-fire, and prevention products for occasional events chronically underperform their ROI math. This is why problem intensity scored 15, not 18.
3. Liability perception: If a dealer follows LightCall’s recommendation and still loses an arbitration, they will blame the tool. Positioning must be decision support with a retained evidence file, never a guarantee — and the defense-file feature is what makes that positioning honest rather than a disclaimer.
4. Rules maintenance tail: NAAA updates guidelines and each auction runs house rules. If the corpus goes stale the product becomes actively harmful. This is an ongoing operating cost, not a one-time build.
5. Capture-quality dependency: The output is only as good as the photos a lot porter takes at 4pm in the rain. Garbage in, wrong announcement out.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder paired with an ex-wholesale/used-car manager
Time to revenue:        6–10 weeks (manual service first, software behind it)
Capital to launch:      $15–25K
Top 3 assumptions to validate first:
  1. Median seller arbitration frequency ≥3/year — 25 used-car manager interviews with actual counts
  2. Experienced managers disagree on the light-and-announce call — blind test 15 managers on 10 real units
  3. Photo + OBD capture catches the defects that actually get arbitrated — retrospective check on 30 real claims
Kill criteria:
  - Abandon if median dealer reports <3 arbitrations/year across 25 interviews — the event is too rare to price
  - Abandon if the blind test shows >80% agreement among experienced managers — the decision isn't hard enough to sell
  - Abandon if ACV or Cox ships a seller-side announcement assistant before v1
  - Abandon if fewer than 5 of 40 manual-service pilots convert to paid at $199/mo

15. Next step — 1-week validation sprint

• Day 1–2: Pull the June 1, 2026 NAAA guidelines and the arbitration policies of Manheim, ADESA, and three regional auctions. Build the announcement decision table by hand for the 20 most common defect categories. If I can’t produce a defensible table in two days, the rules corpus is harder than I think and the timeline is wrong.
• Day 3–4: Call 25 used-car managers from state IADA directories. Two questions only: how many arbitrations did you eat in the last 12 months and what did each cost, and who decides the light on a given car. Simultaneously run the blind test — 10 real units with defect lists, 15 managers, compare their calls.
• Day 5: Decide. Go if median reported arbitrations ≥3/year AND the blind test shows managers disagreeing on ≥4 of 10 units. No-go if either fails.

The falsifiable result is the blind test. If fifteen experienced used-car managers independently make the same light-and-announce call on the same ten cars, there is no product here — the decision is tacit knowledge that already works, and I should walk away rather than build a tool that tells experts what they already know.