Heavy Equipment Credit Risk & Commercial Scorecard Engine

01. Problem Architecture & Commercial Underwriting Bottlenecks

Financing high-value capital assets—such as 20-ton hydraulic excavators, off-highway dump trucks, and track-type bulldozers—presents fundamentally distinct credit risk characteristics compared to retail consumer lending. Single contract exposures routinely range between $80,000 and $500,000 USD (equivalent to IDR 1.2B to 7.5B), creating high loss concentration per default event:

[Contractor Application]
         │
         ▼
[Credit Bureau Ingestion] ────► [Delinquency Anomaly Guard (96/98 Isolation)]
         │
         ▼
[Financial 5C Evaluation] ────► [DSCR >= 1.15x] & [DER <= 2.00x] & [CCR >= 120%]
         │
         ▼
[Machine Learning Engine] ────► [HistGradientBoosting (AUC 0.8688) & Logistic (AUC 0.8615)]
         │
         ▼
[Basel II Scorecard Scale] ───► [Points = Offset + Factor * ln(Odds)] ──► Range [300 - 850]
         │
         ▼
[Macroeconomic Overlay]   ────► [ESDM HBA Coal Price Shock Multiplier]
         │
         ▼
[IFRS 9 Expected Loss]    ────► EL = PD * LGD * EAD ──► [Credit Committee Memo (NAK)]

In volatile emerging-market resource economies, financiers face three systemic failure modes:

1. Commodity Price Vulnerability: Contractor operating margins contract sharply when thermal coal prices (ESDM HBA) or palm oil benchmark prices dip, depressing Debt Service Coverage Ratios (DSCR) below the 1.0x break-even point.
2. Hidden Leverage in Stated Income: Retail credit bureaus frequently report raw monthly debt obligations as debt ratios when income is omitted, introducing massive distortion if left unnormalized.
3. Depreciation vs Amortization Mismatch: Heavy machinery experiences steep initial operational wear (~15% annual market depreciation). Without front-loaded down payments (≥ 20%) and linear principal amortization, asset recovery values plunge below outstanding debt balances during foreclosure.

02. Dataset Integrity & Anomaly Guard (GiveMeSomeCredit 150k Cohort)

The underwriting pipeline operates on the complete 150,000 borrower records from GiveMeSomeCredit, evaluating binary default risk (SeriousDlqin2yrs = 90+ days past due or charge-off within 24 months). Exploratory audits revealed two severe data anomalies requiring robust feature guards:

The 96/98 Delinquency Error Code Cohort

A subset of 269 borrowers exhibited extreme delinquency counts of 96 or 98 across all past-due columns (NumberOfTime30-59DaysPastDueNotWorse, NumberOfTime60-89DaysPastDueNotWorse, NumberOfTimes90DaysLate). Rather than legitimate incident counts, these represent credit bureau processing exception codes.

Because the default rate for this anomaly group is 54.65% (more than 8 times the population baseline), dropping these rows would erase critical credit risk signal. The pipeline creates a dedicated binary feature is_delinquency_error_code = 1 while capping the numeric counters at the 99th percentile (5).

!Credit Bureau 96/98 Anomaly Default Multiplier

Revolving Utilization Distribution:
[ 0.00 - 0.30 ]  ■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ (2.22% Default Rate)
[ 0.30 - 0.70 ]  ■■■■■■■ (7.40% Default Rate)
[ 0.70 - 1.00 ]  ■■■■■■■■■■■■■■■■■ (17.72% Default Rate)
[    > 1.00   ]  ■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■ (37.25% Default Rate)

03. The 5C Credit Appraisal Mathematical Framework

Institutional commercial underwriting synthesizes qualitative business governance with rigorous quantitative ratio analysis across the Five Cs of Credit:

A. Character (Biro Kredit & Integritas Rekam Jejak)

Measures willingness to pay via past delinquency severity. A composite delinquency penalty index isolates repeated short-term arrears from chronic non-performance:

B. Capacity (Kapasitas Pembayaran Arus Kas)

Evaluates whether project revenues sufficiently service debt amortizations. The primary hurdle is the Debt Service Coverage Ratio (DSCR):

*Target Standard*: DSCR ≥ 1.15 × under baseline operations; DSCR ≥ 1.00 × under stress testing.

C. Capital (Struktur Permodalan & Leverage)

Measures balance sheet solvency and owner equity skin-in-the-game via the Debt to Equity Ratio (DER):

*Target Standard*: DER ≤ 2.00 × for mining and agro contractors; hard stop at 2.50 ×.

D. Collateral (Agresi Depresiasi & Perlindungan Agunan)

Guarantees secondary asset recovery. Collateral Coverage Ratio (CCR) is calibrated at origination (Month = 0) and monitored over the 36-month amortization window:

Assuming a standard 15% annual declining balance equipment depreciation rate:

E. Condition (Kondisi Makroekonomi & Siklus Komoditas)

Stresses debtor cash flows against volatility in the Harga Batubara Acuan (ESDM HBA) and Crude Palm Oil (CPO) benchmark indices.

04. Machine Learning Tournament & Basel II Scorecard Calibration

Two distinct model families were trained on an 80:20 stratified split (120,000 train records, 30,000 test records) to predict 24-month delinquency:

!ROC Curves of Credit Scoring Models

!Feature Importance Weights

!Scorecard Probability Distribution

Standardized Logistic Regression Feature Coefficients:
utilization_winsorized          [ +0.7112 ] ■■■■■■■■■■■■■■■■ (Risk Escalator)
NumberOfTimes90DaysLate_capped  [ +0.4579 ] ■■■■■■■■■■ (Severe Delinquency)
NumberOfTime60-89DPD_capped     [ +0.4265 ] ■■■■■■■■■ (Mid-Term Delinquency)
NumberOfTime30-59DPD_capped     [ +0.4081 ] ■■■■■■■■■ (Early Delinquency)
delinquency_penalty_index       [ +0.3324 ] ■■■■■■■ (Composite Score)
age_cleaned                     [ -0.2746 ] ■■■■■■ (Risk Reducer)
DSCR_proxy                      [ -0.1102 ] ■■■ (Cash Flow Buffer)

Basel II Scorecard Point Scaling Engine

To bridge machine learning probabilities with credit committee workflows, raw default probabilities PD are mapped to a traditional 300 to 850 credit scorecard scale:

where Odds = (1 - PD / PD). Enforcing a Base Score of 600 at 50:1 Odds with Points to Double Odds (PDO) = 20:

05. Macroeconomic Commodity Overlay & Sector Sensitivity Stress Test

Debtor performance in natural resource equipment financing is heavily coupled to commodity cycles. The engine incorporates an empirical macroeconomic sensitivity overlay using historical ESDM Harga Batubara Acuan (HBA) pricing data from 2021 through 2024:

Historical ESDM HBA Coal Price vs Heavy Equipment NPL Correlation:
2022-Q2 (HBA $310.8/MT) ──► Mining Equipment NPL: 1.2% (Historical Low)
2023-Q2 (HBA $190.2/MT) ──► Mining Equipment NPL: 2.4% (Normalizing)
2024-Q4 (HBA $115.0/MT) ──► Mining Equipment NPL: 5.4% (Stress Incurred)

!ESDM HBA Coal Benchmark vs Heavy Equipment NPL

!Macroeconomic Stress Testing by Sector

Because mining contractors bear high fixed operating expenses (diesel fuel, equipment maintenance, haul road maintenance), a -20% drop in coal prices triggers an average -13% contraction in net operating cash flows, reducing debt service coverage and escalating default risk by 1.75 ×.

06. IFRS 9 Expected Loss & Secondary Asset Liquidation Economics

Under the IFRS 9 / PSAK 71 accounting framework, credit provisions are calculated via the tripartite Expected Loss (EL) model:

Where:

• PD (Probability of Default): 12-month or lifetime probability derived from the calibrated Basel scorecard.
• LGD (Loss Given Default): Unrecovered exposure post-collateral liquidation, factoring in secondary market liquidity haircuts:
• Excavator 20-Ton: 25.0% LGD (High secondary market liquidity; active inter-island resale demand).
• Heavy Dump Truck: 35.0% LGD (Accelerated chassis fatigue and tire depreciation).
• Bulldozer D85 Track: 35.0% LGD (Undercarriage wear and higher overhaul logistics cost).
• Wheel Loader 3.0 m³: 30.0% LGD (Standard agricultural and batching plant utility).
• EAD (Exposure at Default): Outstanding principal plus accrued interest at default.

For a prime four-unit excavator facility ($320,000 USD / IDR 4.8B exposure) with PD = 1.82% and LGD = 25.0%:

07. Production Implementation & Credit Committee Workflow Integration

The engine is engineered as a zero-dependency, type-safe TypeScript underwriting library (`lib/credit-risk.ts`) integrated into an interactive web cockpit (`components/CreditAnalystStudio.tsx`) and validated by automated regression tests (`lib/credit-risk.test.ts`).

The automated workflow transforms credit origination:

1. Instant Ratio Screening: As underwriting analysts input balance sheet figures, DSCR, DER, and CCR Day 0 compliance flags update in real time.
2. Dynamic Risk-Based Pricing: Rather than flat interest margins, interest rates scale automatically with debtor scorecard rating tiers (10.75% for AAA up to 14.50% for BBB).
3. Automated Credit Memorandum Generation: Generates compliant Nota Analisis Kredit (NAK) summaries ready for instant review and export by Credit Committee executives.