Applicant Risk Assessment

Everything hangs off a single applicant object: creditScore, latePayments, creditHistoryMonths, employmentMonths, incomeVerification, bankAccountStanding, debtToIncomeRatio, and outstandingLoans feed the three scoring tables.

Credit strength earns up to 30 points when all three columns line up: a creditScore above 750 with fewer than 2 latePayments and more than 60 months of history takes the top row, the [650..750] band with under 4 late payments and over 36 months scores 20, [600..650) scores 10, and the blank fallback row scores 0. First hit ordering means an applicant has to clear every column in a row to collect its points.

The bands roughly track familiar consumer credit tiers, with 750 marking prime territory and 650 the near-prime boundary, so the breakpoints read naturally to underwriters even though the point values are template choices. Requiring history length alongside the score matters because a thin file can show a high number with little evidence behind it, and credit-based insurance scoring genuinely relies on these same ingredients where it is permitted.

Stability points demand agreement across three columns: more than 60 employmentMonths with 'complete' incomeVerification and 'good' bankAccountStanding scores 25, the same qualitative pair over 24 months scores 20, 'partial' verification with 'fair' standing past 12 months scores 15, any tenure over 6 months salvages 5 points regardless of the other columns, and the fallback gives 0.

Employment tenure is the classic proxy for income continuity, and letting documented income outrank stated income mirrors the verification tiers lenders and underwriters actually use. The 6, 12, 24, and 60 month breakpoints are sensible template values rather than fixed industry standards, chosen so each step upgrades both duration and documentation together.

Debt burden is scored from applicant.debtToIncomeRatio paired with outstandingLoans: under 0.2 with fewer than 2 loans earns 25 points, under 0.4 with fewer than 3 earns 15, under 0.6 with fewer than 5 earns 5, and the empty row gives 0. Because matching is first hit, a low ratio with many loans slides down to whichever row it fully satisfies.

These breakpoints echo lending practice, where ratios around a third of income are treated as comfortable and anything approaching half as stretched, so rewarding sub-0.2 borrowers most is realistic. Counting loans separately catches applicants who keep the ratio low across many small obligations, a pattern that still signals payment complexity.

Component sums come together as totalRiskScore, the sum of scores.creditHistory, scores.incomeStability, and scores.debtToIncome, for a maximum of 80. The negativeFactors filter collects every component that scored 0 and negativeFactorsCount measures how many, so the classifier can tell a mediocre-everywhere applicant apart from one with a single failed dimension.

Routing runs on two signals at once: totalRiskScore >= 60 and negativeFactorsCount == 0 takes the low-risk branch, totalRiskScore >= 30 or (totalRiskScore >= 20 and negativeFactorsCount < 2) goes to medium, and the default statement sweeps the rest into high risk. Making the top tier require zero failed components keeps a file with one blank dimension out of auto-approval no matter how strong the rest of it is.

Files that reach this branch get riskCategory 'low', approvalStatus 'auto-approved', and an interestRateModifier of -0.5, a small pricing credit. Auto-approving only the cleanest tier is the straight-through processing pattern that keeps underwriters off routine applications.

The middle branch sets approvalStatus to 'manual-review' with a neutral interestRateModifier of 0, so pricing stays standard while a person decides. Holding the modifier flat until an underwriter has looked is a conservative default for borderline scores.

Anything falling through the default statement lands here: riskCategory 'high', approvalStatus 'additional-verification', and a 1.5 point interestRateModifier. Requesting more documentation instead of declining outright preserves the sale, while the surcharge prices the extra uncertainty in the meantime.