LineHound — supplement engine for collision repair shops

1. One-liner

Flags every billable operation the insurer left off a collision estimate — with the OEM citation that gets it paid.

2. Trend signal — why now?

Three things are converging in collision repair right now, and they all push money toward the shop that knows how to ask for it.

Repairs are getting more complex and more expensive per car, while volume falls. CCC’s 2026 Crash Course data shows repairable claims down 10.4% year-over-year through August 2025, but each repair is harder: the share of DRP estimates including at least one calibration jumped from 26.9% to 35.6% in a single year. Fewer cars, more billable operations per car — exactly the environment where capturing every line item decides whether a shop makes its month.

Insurers are systematically short-paying. Industry reporting is blunt about it: insurers “are increasingly omitting obvious damage and necessary procedures, hoping the vehicle owner won’t get the car repaired or will find a shop that will repair it for the amount the insurance company paid.” Reimbursement lagged labor cost by ~8% in 2023. The original estimate is a floor, not the truth, and the shop has to write a supplement to recover the rest.

The estimators who catch all of it are rare — and AI just got good enough to do the catching. A single missed 0.3-hour operation on 150 repair orders a month is ~$1,890/month, $22,680 a year, on one overlooked line. The “not-included operations” buried in P-pages and OEM repair procedures are precisely the items estimators forget. Reading an estimate, cross-referencing OEM position statements and P-page footnotes, and listing every missing billable op is now a cheap vision-plus-reasoning task that wasn’t reliably possible 12 months ago.

Provenance:

• Signal 1 (Demand): “Shops doing 150 repair orders a month missing just one operation worth three-tenths of an hour on every estimate could lose $1,890 a month or $22,680 a year”; “techs spend time checking P-pages and writing their own supplements because the estimator didn’t know everything to ask for” — https://www.bodyshopbusiness.com/don-t-underestimate-the-importance-of-estimators/ and https://www.autobodynews.com/news/common-estimating-mistakes-could-be-costing-your-shop-money — June 2026
• Signal 2 (Feasibility): OEM position statements and repair procedures (pre/post scan, calibration, corrosion protection) are the citable evidence a supplement needs; centralized at I-CAR RTS and oem1stop; CCC Secure Share API now exposes shop estimate data to third-party apps across 22,000+ facilities — https://rts.i-car.com/oem-information.html and https://www.cccsecureshare.com/Developers — June 2026
• Signal 3 (Economic): Repairs down 10.4% but calibrations on DRP estimates up from 26.9% to 35.6% YoY; reimbursement lagging labor cost ~8%; CCC’s own Smart Audit applies AI to flag line-level outliers for insurers — money is moving to per-job estimate intelligence — https://www.autobodynews.com/news/2025-data-points-to-fewer-claims-more-collision-repair-complexity-in-2026 — June 2026 Category: Workflow automation

3. The opportunity

Every collision estimate is a negotiation, and the shop usually shows up unarmed. The insurer’s adjuster writes an estimate that omits “not-included operations” — feather/prime/block, cover and protect, weld destructive testing, corrosion protection, R&I to access, pre- and post-repair scans, ADAS calibrations. Some of these are forgotten honestly; many are left off deliberately because most shops won’t fight for them. The shop’s defense is one tired estimator’s memory plus a laminated ASA “Not-Included Operations” chart.

The incumbents won’t fix this, and that’s the whole opening. CCC, Mitchell, and Audatex write the estimates — they sell to insurers too, and CCC’s Smart Audit is explicitly an AI tool that flags line items for insurers to reinspect and cut. They have a structural conflict of interest: a tool that tells the shop “you under-wrote this, the carrier owes you $640 more, here’s the OEM citation” works directly against their largest customers. No incumbent ships that. A focused outsider can.

The 10× isn’t a prettier estimating UI. It’s turning a 30-minute manual P-page hunt that most shops skip into a 90-second automated pass that produces a line-by-line, OEM-cited supplement worksheet — on every car, not just the ones the best estimator has time to scrutinize.

4. Target market

• Primary customer: Owner or lead estimator/blueprinter at an independent collision/auto-body shop in the US — single location to small MSO (1–8 bays), $500K–$4M annual revenue, writing 30–150 estimates/month, not buried under a DRP that dictates every line. The estimator is the daily user; the owner signs up because it’s pure recovered margin.
• Why they buy (their words): “Techs shouldn’t have to write their own supplements because the estimator didn’t know everything to ask for.” “If you take the insurance company’s payment based on the preliminary estimate, you are leaving money on the table.” They feel it on every job.
• Rough TAM reasoning: ~32,000 independent body shops in the US (industry ~$76.9B). Even at a $149/mo ACV, the independent segment alone is a ~$57M/yr addressable line before MSOs and adjacent shops (PDR, hail/cat, glass) widen it.
• Why now for them: Falling car counts mean each repair has to be fully captured to hit the month. Calibration and ADAS complexity is exploding the number of legitimate not-included ops — and the dollar value of each one a shop misses.

5. Product sketch (MVP)

• Upload an estimate (PDF, or EMS/BMS export from CCC/Mitchell/Audatex) — or connect once via CCC Secure Share.
• LineHound parses every line and identifies the vehicle, the operations written, and the parts being replaced or R&I’d.
• It returns a missing-operations worksheet: each billable op the estimate likely omits (scan, calibration, corrosion protection, feather/prime/block, cover-and-protect, weld test, R&I-to-access), grouped by confidence.
• Each flagged line ships with its justification citation — the OEM position statement, repair-procedure reference, or P-page “not-included” rule that supports billing it.
• One-click supplement note: copy-ready language the estimator can paste into the supplement or hand the adjuster, line by line.
• Per-job recovery tally and a monthly “money you would have left on the table” report the owner actually reads.
• Vehicle/OEM-aware: knows that this Hyundai requires pre- and post-scan, that this model-year needs a specific calibration after that R&R.

6. AI angle — what’s load-bearing

Remove the AI and there is no product. The core job is reading a messy, semi-structured estimate (often a scanned PDF with shop-specific abbreviations), inferring the actual repair operations and the vehicle, and reasoning over a large, footnote-heavy corpus of OEM procedures and P-page “not-included” rules to decide which billable operations are missing. That’s document understanding plus structured reasoning over a knowledge base — exactly what got cheap and reliable in the last 12 months and was brittle before. The competing manual process is an experienced estimator with crash books, which is precisely the scarce, expensive resource shops can’t staff. AI isn’t a chatbot bolted onto an estimating tool here; it’s the estimator’s missing second brain.

7. Localization angle (if any)

N/A — this is a US-first play, and that’s deliberate. The wedge depends on the US insurer-vs-shop supplement dynamic, OEM position statements as billing evidence, P-page conventions, and the CCC/Mitchell/Audatex estimating duopoly-plus. Those don’t transfer cleanly to markets without the same DRP/supplement structure. A later UK/Canada cut is plausible but is a different ruleset, not a translation. Forcing a localization angle here would weaken the idea.

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

• Pricing: $149/mo flat for a single shop (unlimited estimates); $99/mo/location for MSOs. A usage/per-estimate tier ($1–2/estimate) for very low-volume shops to remove the adoption barrier.
• ACV: ~$1,800/year single-shop average; higher for multi-location.
• Rough math to $1M ARR: 560 shops × $149/mo × 12 ≈ $1.0M. That’s under 2% of the ~32,000 independent shops.
• Rough math to $5M ARR: ~2,800 shops (≈9% of independents) at the same ACV, or 1,800 shops plus MSO expansion and a per-estimate API tier sold to estimate-review/consulting firms.
• Expansion path: Start at per-shop supplement capture → add OEM-procedure attachment packets (printed evidence for the file) → labor-rate justification add-on → a benchmarking layer (“shops like yours recover $X more per RO”) that compounds with data. ACV grows as the shop trusts it on more of the negotiation, not just the line list.

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

• Trade media + association beachhead. BodyShop Business, Autobody News, FenderBender, and CollisionWeek already run the exact “you’re leaving money on the table” articles this product solves. Pitch a guest piece / sponsored teardown: “We ran 50 real estimates through LineHound — here’s the average $ missing per car.” That single number is the entire sales pitch.
• Facebook groups and forums where estimators live. Collision-repair owner/estimator Facebook groups and forums are large, active, and full of exactly this complaint. Offer a free “score my estimate” — upload one redacted estimate, get the missing-ops list back. The output sells itself because it’s denominated in dollars.
• Free estimate teardown as the top of funnel. Scrape/compile the ~32,000 independent shops (state DMV/repair-facility registries, Google Maps), send a personalized one-page teardown of a sample estimate for their make-mix, expect a few percent to book a demo. The artifact is the demo.
• CCC Secure Share marketplace listing. Once integrated, LineHound appears in a marketplace reaching 22,000+ repair facilities with a single connection — inbound distribution most outsiders can’t get to.
• Estimate-review consultants and “Diminished Value”/supplement consultants already do this manually for a fee; sell them the per-estimate API tier and let them resell.

10. Build complexity — justification

Medium. The web app, file parsing, and LLM reasoning are off-the-shelf. The real work is the OEM/P-page knowledge base — assembling and structuring position statements, not-included rules, and procedure references into something the model can cite accurately, then tuning precision so the tool doesn’t cry wolf (a false “you missed this” that an adjuster swats down burns trust fast). EMS/BMS parsing and the CCC Secure Share integration are documented but real integration work. A founder with collision-estimating domain expertise plus one engineer ships a credible v1 in ~3–4 months; the knowledge base keeps improving after launch.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helping a shop bill for legitimately-required operations with OEM citations is standard practice; no regulated activity.
Ethical — no harm / dark patterns	✅	The tool flags legitimate, required operations with documentation. It’s anti-fraud, not pro-padding — and should be tuned to defensible lines only.
Market exists (evidence above)	✅	32,000 independent shops, documented systemic short-pay, quantified per-job losses.
1–5 person team can build this	✅	Founder + 1–2 engineers; knowledge base is the long pole, not headcount.
Launchable with <$50K / ₹40L	✅	Off-the-shelf stack; main cost is domain time assembling the OEM corpus.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Felt on every single repair; directly costs money; quantified at ~$22K/yr from one missed op. Just shy of hair-on-fire because many shops have built tolerance/workarounds.
Demand evidence	15	12/20→12/15	Multiple independent industry sources quantifying the loss; incumbents (CCC Smart Audit) building the insurer side proves the value of the data. Direct “shops will pay for the shop-side version” still needs a paid pilot.
Build feasibility	15	11/15	App and AI off-the-shelf; OEM/P-page knowledge base and precision tuning are the real work. ~3–4 months to v1.
Distribution clarity	15	11/15	Named trade media, named forums/groups, a free teardown artifact denominated in dollars, plus the Secure Share marketplace. Conversion math still unproven.
Revenue mechanics	15	12/15	Pricing trivially justified by ROI; $1M needs <2% of independents. Churn risk if perceived value fades once the obvious misses are captured.
Time to first revenue	10	8/10	The teardown-to-paid funnel is short; owners pay fast when the output is in dollars. Needs the knowledge base solid enough to be trustworthy on day one.
Defensibility	10	5/10	Copyable in concept. Moat is the accuracy of the OEM/P-page corpus, accumulating per-shop benchmark data, and workflow lock-in — a 6–12 month head start, not a wall.
Total	100	75/100

13. Qualitative modifiers

Founder-fit tags

domain-expertise-required · technical-heavy

This idea lives or dies on collision-estimating credibility. A founder who has written estimates and fought supplements — or a tight pairing with someone who has — will get the precision right and earn trust in the forums. A pure-tech team will ship something that looks right and quietly cries wolf, and the market will know within a week.

Key assumptions to validate (3–5)

1. Assumption: Independent shops will pay ~$149/mo for recovered supplement dollars. How to test: Run 40–50 real estimates through a manual+AI prototype, show shops the per-job $ missed, and pre-sell paid pilots. Need ≥15% of shown-the-number shops to commit.
2. Assumption: The tool can flag missing ops with high enough precision that adjusters accept them (low false-positive rate). How to test: Have 3–5 veteran estimators grade 100 LineHound outputs against ground truth; target <10% “this would get laughed out of the negotiation” rate.
3. Assumption: The OEM/P-page corpus is assemble-able and licensable/usable at acceptable cost. How to test: Audit availability and terms of OEM position statements, I-CAR RTS, P-page access, and ALLDATA/OEM procedure sourcing before committing.
4. Assumption: Shops can/will get estimates into the tool (upload friction is low). How to test: Time real shops doing PDF upload vs EMS export vs Secure Share; confirm at least one path is <60 seconds.

Risk flags

1. Platform dependency: CCC controls Secure Share and the dominant estimating platform; they could restrict third-party access or ship a competing shop-side feature (though the insurer conflict makes the latter unlikely). Keep PDF/EMS upload as a platform-independent fallback.
2. Accuracy/trust risk: False positives are existential. One adjuster swatting down a bogus LineHound line and the estimator stops trusting it. Precision over recall, tuned to defensible lines only.
3. Value-decay / churn: Once a shop’s estimators internalize the obvious misses, perceived value can drop. Counter with the harder calibration/OEM-procedure long tail and the benchmarking/reporting layer that keeps surfacing new money.
4. Insurer pushback: Carriers may resist a tool that systematically raises supplement volume; unlikely to be a legal issue (the ops are legitimately required) but could create DRP friction for shops that depend on a carrier relationship.

14. Structured verdict

Score:                  75/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Collision-estimating domain expert paired with a technical co-founder (or a technical founder with a paid estimator advisor from day one)
Time to revenue:        6–10 weeks (paid pilots off the teardown funnel)
Capital to launch:      $15–35K (mostly domain time assembling the OEM/P-page corpus)
Top 3 assumptions to validate first:
  1. Shops pay ~$149/mo for recovered dollars — pre-sell ≥15% of shops shown their per-job $ missed
  2. Flagging precision is adjuster-proof — veteran estimators grade outputs, <10% indefensible
  3. OEM/P-page corpus is usable at acceptable cost — audit sourcing/licensing before building
Kill criteria:
  - Abandon if <15% of 40 shops shown their per-job missed-dollars will commit to a paid pilot
  - Abandon if veteran estimators rate >25% of flagged lines as indefensible (false positives kill trust)
  - Abandon if OEM procedure / P-page data cannot be sourced at a cost compatible with $149/mo pricing

15. Next step — 1-week validation sprint

• Day 1–2: Collect 30–50 real (redacted) collision estimates — from forum volunteers, a friendly shop, or estimate samples. Hand-build the “missing ops” answer key for 20 of them with a veteran estimator.
• Day 3–4: Stand up a rough pipeline (parse estimate → reason over a starter OEM/P-page knowledge base → output missing-ops list with citations). Run all 50 estimates; measure precision against the answer key.
• Day 5: Show 30 shop owners/estimators their own estimate’s per-job missed dollars and ask them to commit to a paid pilot.

Falsifiable outcome: Go only if (a) precision on flagged lines clears ~75% against the veteran-built key, AND (b) ≥15% of the 30 shops shown their number commit to a paid pilot. Miss either and it’s a no-go, not a “they seemed interested.”