Claims Valuation — Insurance Total-Loss & Settlement Pricing

Collect from user:

• VIN (required)
• Current odometer reading (required)
• Pre-loss condition: Clean, Average, Rough (required)
• Date of loss (optional, defaults to today)
• Any pre-existing damage or modifications

Call mcp__marketcheck__decode_vin_neovin with the VIN to get exact specs: year, make, model, trim, body type, drivetrain, engine, transmission, original MSRP. → Extract only: year, make, model, trim, body_type, drivetrain, engine, transmission, MSRP. Discard full response.

Step 2 — Fair Market Value (FMV) determination

Make THREE pricing calls:

1. mcp__marketcheck__predict_price_with_comparables with VIN, miles, ZIP, dealer_type=franchise → Franchise retail FMV
2. mcp__marketcheck__predict_price_with_comparables with VIN, miles, ZIP, dealer_type=independent → Independent retail FMV
3. If vehicle was CPO (is_certified in history or user states): additional call with is_certified=true → Extract only: predicted_price, comp count per call. Discard full response.

Pre-loss FMV = average of franchise and independent predicted prices, adjusted by condition:

• Clean: use the higher of the two predictions
• Average: use the average
• Rough: use the lower, minus 5%

Step 3 — Comparable evidence (wide radius)

Pull active retail comparables:

• mcp__marketcheck__search_active_cars with YMMT, ZIP, radius=100, miles_range=<odo-15000>-<odo+15000>, car_type=used, sort_by=price, sort_order=asc, rows=20 → Extract only: VIN, price, miles, dealer_name, distance, dom per listing. Discard full response.

Pull sold transaction evidence (strongest for disputes):

• mcp__marketcheck__search_past_90_days with same YMMT filters, sold=true → Extract only: VIN, sold_price, miles, dealer_name, sale_date per listing. Discard full response.

Step 4 — Total-loss determination

Calculate:

• Repair cost threshold = FMV x total_loss_threshold_pct (default 75%)
• If estimated repair cost > threshold → TOTAL LOSS
• If not provided, just present the threshold: "This vehicle is a total loss if repair costs exceed $XX,XXX (75% of FMV)"

Step 5 — Settlement range

Calculate three tiers:

• Low settlement = 25th percentile of sold transaction prices (condition-adjusted)
• Mid settlement = Pre-loss FMV from Step 2
• High settlement = 75th percentile of sold transaction prices

Step 6 — Salvage value estimate

Salvage value is typically 15-25% of pre-loss FMV depending on damage severity:

• Minor (cosmetic): 25% of FMV
• Moderate (mechanical): 20% of FMV
• Severe (structural/flood): 15% of FMV
• If user provides actual salvage bid, use that instead

Net claim cost = Settlement offer - Salvage value

Output

Present: vehicle ID summary, pre-loss FMV table (franchise/independent/condition-adjusted), total-loss threshold and determination, settlement range (low/mid/high), salvage estimate and net claim cost, comparable evidence tables (active + sold with VIN/price/miles/dealer), methodology notes and caveats.

Workflow: Batch Claims Processing

For multiple VINs (e.g., hail damage event, flood):

1. Accept list of VINs with miles and condition
2. Use the insurer:portfolio-scanner agent to process each VIN
3. Present summary: total claim exposure, average FMV, total-loss count, salvage estimate
4. Ranked table of all vehicles with FMV and settlement recommendation

Workflow: Diminished Value

For vehicles that were repaired (not total-loss) and the claimant wants diminished value:

1. Get pre-loss FMV (same as Steps 1-2 above)
2. Estimate post-repair diminished value: typically 10-25% of FMV depending on repair severity
3. Show: "Pre-loss FMV: $XX,XXX → Post-repair diminished value: $X,XXX-$X,XXX"

Important Notes

• US-only: Requires predict_price_with_comparables and search_past_90_days.
• Always cite specific comparable VINs — adjusters need to defend valuations.
• Sold transaction data is the strongest evidence in settlement disputes.
• The 75% total-loss threshold is a common default — actual thresholds vary by state and insurer. Use profile value if available.
• For disputed settlements, recommend expanding the search radius to 150-200 miles and including older sold data.