lending-fit-analyzer

Date anchor: Today's date comes from the # currentDate system context. Compute ALL relative dates from it. Example: if today = 2026-03-14, then "prior month" = 2026-02-01 to 2026-02-28, "current month" (most recent complete) = February 2026, "three months ago" = December 2025. Never use training-data dates.

Lending Fit Analyzer — Overlay Lending Criteria on a Dealer's Lot

Profile

Load the marketcheck-profile.md project memory file if exists. Extract: price_range_min, price_range_max, preferred_year_range, max_mileage, approved_makes, approved_segments, ltv_max_pct, country, zip, radius. If missing, ask minimum fields. US: search_active_cars, predict_price_with_comparables. UK: search_uk_active_cars only (no LTV check — skip predict calls). Confirm: "Using profile: [company], [lending_type]". All preference values from profile — do not re-ask.

User Context

Lender sales rep wants to precisely quantify how much of a dealer's lot they can finance. Used to prepare proposals, justify partnerships, and identify gap opportunities.

Gotchas

1. LTV direction matters. LTV = listed_price / predicted_price x 100. Values OVER 100% mean the dealer is listing above market value (higher lender risk). Values UNDER 100% mean the dealer is listing below market (lower risk, easier to advance on). Do not invert this ratio.
2. UK profiles — no LTV analysis. predict_price_with_comparables is US-only. For UK dealers, skip step 3 (LTV analysis) entirely and present coverage % only. Note "LTV analysis unavailable for UK market" in the output.
3. Sampling bias in LTV spot-check. The 10-unit sample (3 cheapest, 4 median, 3 most expensive) is designed to capture the range, but may miss outliers. Always note the sample size and that LTV estimates are based on a representative sample, not a full-lot appraisal.
4. Make/segment filters are AND conditions. If both approved_makes and approved_segments are set, the search requires BOTH to match. This can dramatically reduce the matching count. Report the filter combination used so the user understands why coverage might be low.
5. Gap analysis should be actionable. Saying "200 units are outside price range" is less useful than "expanding your price ceiling from $55K to $65K would add 45 units (12% coverage gain)." Always quantify what loosening one constraint would yield.

Workflow: Full Fit Analysis

1. Get total dealer inventory — Call mcp__marketcheck__search_active_cars with dealer_id (or source), car_type=used, rows=0, stats=price,miles,dom, facets=make|0|20|1,body_type|0|10|1,year|0|10|1. → Extract only: total_count, stats (avg_price, median, min, max), facets. Discard full response.

2. Get matching inventory — Call mcp__marketcheck__search_active_cars with same dealer, price_range=[min]-[max], year=[year_range], miles_range=0-[max_mileage], rows=0, stats=price,miles, facets=make|0|20|1,body_type|0|10|1. If approved_makes is set, add make=[comma-separated]. If approved_segments is set, add body_type=[comma-separated]. → Extract only: matching_count, stats, facets. Discard full response.

3. LTV analysis — For a representative sample (up to 10 units: 3 cheapest, 4 median, 3 most expensive from matching set):

   • Call mcp__marketcheck__search_active_cars with same dealer + criteria, sort_by=price, sort_order=asc, rows=3 for cheap end.
   • Call mcp__marketcheck__search_active_cars with same dealer + criteria, sort_by=price, sort_order=desc, rows=3 for expensive end.
   • Call mcp__marketcheck__search_active_cars with same dealer + criteria, rows=4, start=[matching_count/2 - 2] for middle range. → Extract only per listing: vin, listed_price, miles, year, make, model.

   For each sampled VIN, call mcp__marketcheck__predict_price_with_comparables with vin=[VIN], miles=[miles], zip=[dealer_zip], dealer_type=[dealer's type]. → Extract only per VIN: predicted_price, comp_count. Calculate LTV = listed_price / predicted_price x 100. If comp_count < 5, flag as "low confidence." Discard full responses.

4. Calculate fit metrics:

   • Coverage % = matching_count / total_count × 100
   • LTV Distribution from sampled units:
     • Under 100% (under-advanced): count and %
     • 100-110% (standard): count and %
     • 110-[ltv_max]% (elevated): count and %
     • Over [ltv_max]% (exceeds limit): count and %
   • Estimated Portfolio Value = matching_count × avg_matching_price × estimated_advance_rate (based on avg LTV)
   • Monthly Volume Estimate = matching_count × (1 / avg_dom × 30) × penetration (20%)

5. Gap analysis — What's NOT covered and why:

   • Units outside price range: count and avg price
   • Units too old: count and avg year
   • Units too high mileage: count and avg miles
   • Units with wrong make: count (if make filter applied)
   • Suggest: "If you expanded price range to $[X], you'd pick up [Y] more units"

Workflow: Quick Coverage Check

When user says "quick fit check" — run steps 1-2 only, return coverage % and matching count. Skip LTV sampling.

Output Template

── Lending Fit Report: [Dealer Name] ── [Date] ─────────────

COVERAGE SUMMARY
Total Used Units:       [X]
Matching Units:         [Y]
Coverage:               XX.X%
Fit Classification:     [STRONG (>60%) / MODERATE (30-60%) / LIGHT (<30%)]

CRITERIA APPLIED
Price:    $[min] - $[max]
Year:     [min_year] - [max_year]
Mileage:  0 - [max_mileage]
Makes:    [list or "all"]
Segments: [list or "all"]

PRICING COMPARISON
                All Inventory    Matching Only
Avg Price:      $XX,XXX          $XX,XXX
Median Price:   $XX,XXX          $XX,XXX

LTV ANALYSIS (based on [N]-unit sample)
Under 100% (under-advanced):    X units (XX%) — low risk
100-110% (standard):            X units (XX%) — normal
110-[max]% (elevated):          X units (XX%) — monitor
Over [max]% (exceeds limit):    X units (XX%) — DECLINE

Avg Estimated LTV:              XXX.X%
LTV Range:                      XX% - XXX%

PORTFOLIO ESTIMATE
Estimated Portfolio Value:       $X,XXX,XXX (matching units x avg price x [advance_rate]%)
Monthly Volume Estimate:         XX units ([matching] x 30/[avg_dom] x 20% penetration)

GAP ANALYSIS — Why [Z] units don't qualify:
- Outside price range:  [X] units (avg $XX,XXX) → expanding to $[Y] adds [N] units
- Too old:              [X] units (avg [year]) → expanding to [year] adds [N] units
- Too high mileage:     [X] units (avg XX,XXX mi) → expanding to [mi] adds [N] units
- Wrong make:           [X] units (if make filter applied)

MATCHING INVENTORY PROFILE
Top Makes:    [Make1] XX%, [Make2] XX%, [Make3] XX%
Body Types:   [Type1] XX%, [Type2] XX%
Year Range:   [oldest]-[newest]

RECOMMENDATION
This dealer is a [STRONG/MODERATE/LIGHT] fit. [X]% of their lot qualifies.
Recommended approach: [specific action based on data].

Source: MarketCheck inventory data, [Date].

Output

Fit report: Total Units, Matching Units, Coverage %, Avg Price (all), Avg Price (matching). LTV Distribution chart: Under 100%, 100-110%, 110-120%, Over 120%. Estimated portfolio value and monthly volume. Gap breakdown: why units don't qualify, with expansion suggestions. Matching inventory profile: top makes, body types, year range. Recommendation: "This dealer is a [STRONG/MODERATE/LIGHT] fit. [X]% of their lot qualifies. Recommended approach: [specific]."

Self-Check (before presenting to user)

1. Coverage % math is correct? matching_units / total_units x 100. Verify matching_units comes from the filtered search and total_units from the unfiltered search. They should not be equal unless coverage is 100%.
2. LTV ratio direction is correct? LTV = listed_price / predicted_price x 100. Over 100% = listing above market (higher risk). Under 100% = listing below market. Verify you did not invert this.
3. Gap analysis numbers add up? The sum of "outside price range" + "too old" + "too high mileage" + "wrong make" should approximately equal (total_units - matching_units). They may not sum exactly due to overlap, but should be in the right ballpark. Note overlaps if significant.
4. Expansion suggestions are quantified? Each gap category should include "expanding [criteria] to [value] would add [N] units." This requires additional API calls — if not made, note as "estimate" or omit.
5. LTV sample is representative? Confirm you sampled from cheap, median, and expensive ends. If the matching set has fewer than 10 units, sample all of them.
6. Fit classification matches thresholds? STRONG (>60%), MODERATE (30-60%), LIGHT (<30%). Verify the label matches the actual coverage %.