6. Underwriting

This section is the platform’s credit policy in spec form. Every rule, every threshold, every scorecard is illustrative — adapt to your appetite. The architectural lesson is to make these configurable rules in the engine, not hard-coded in service logic.

Sub-pages

Page	Covers
6.1 Eligibility and bureau rules	Hard cuts, bureau-derived scorecard
6.2 GST and bank-statement rules	Cash-flow scorecard inputs
6.3 Cash-flow and Tally rules	Operating cash-flow, working-capital cycle, accounting signals
6.4 Fraud and deviation rules	Fraud scorecard, deviation matrix
6.5 Exposure, pricing, tenure grids	Risk grade → price; ticket grid; tenure grid; exposure caps
6.6 Sample rules library	A library of 40+ example rules in the required format
6.7 MVP vs advanced scorecard	Minimum viable vs production-grade scorecards; ML roadmap

Rule template

Every rule in this spec follows this template:

Rule name:
Purpose:
Data source:
Logic:
Threshold:
Pass/fail/deviation:
System action:
Manual review condition:
Audit evidence:

Example — see 6.6 for the full library.

Decision shape

For each application, the engine produces:

• Outcome: APPROVE, DECLINE, REFER (manual), DEFER (more data needed).
• Risk grade: A1, A2, B1, B2, C1, C2, D (illustrative — adjust to your taxonomy).
• Recommended sanction amount, tenure, rate, fee structure.
• Reasons — full list of rule outcomes contributing to the decision.
• Conditions — pre-conditions the borrower must satisfy before drawdown (e.g., assign specific collection account; submit specific document).

Reading order

1. Read 6.1 first — sets the foundation.
2. 6.2 and 6.3 build the core cash-flow scoring.
3. 6.4 layers risk.
4. 6.5 converts decision into terms.
5. 6.6 is the worked example library.
6. 6.7 covers the journey from rule-based to model-based.