BSA Signal Profiler: Map Your Underwriting Policy Against Modern Bank Statement Analyzer Output

Three reasons. First, legacy credit policy was written when bank statement analyzers produced 8-12 ratio-style signals (AMB, bounce count, EMI count, salary date) and the policy hard-coded those exact fields into the scorecard. Second, policy refresh cycles run 18-24 months at most NBFCs, so even when the vendor adds 25 more signals the policy team has not caught up. Third, the credit team often does not know which incremental signals the vendor exposes — the product manager and the policy owner have not had a structured conversation about the full signal catalogue. The result is a 30-40% catalogue utilisation rate, which means lenders are paying for signal coverage they do not consume.

The four-layer synthetic financial stack — personal-vs-business transaction separation, synthetic P&L, synthetic balance sheet, and synthetic cash flow — is the single most-important addition for any lender doing unsecured MSME above ₹10 lakh ticket. MSME borrowers usually do not have audited financials, the GST return is the only formal disclosure, and the sole-proprietor current account mixes personal and business flows. Without synthetic financials the underwriter is effectively guessing. After synthetic financials, the next-highest-impact signals are working capital cycle (sizes the CC limit), GST credit cross-validation (catches under-reporting), and counterparty concentration risk (flags single-OEM dependency).

Rule-based signals are direct measurements off the statement — bounce count, AMB, EMI count, ATM withdrawal frequency. They are deterministic, easy to encode in any scorecard, and have been in production at every BSA vendor for a decade. Synthetic-financial signals are engineered derivations that classify, group, and project transactions into accounting-grade outputs — a derived P&L, a derived cash flow statement, a working-capital cycle. They require multi-pass transaction classification, counterparty grouping, and personal-vs-business segregation logic. Rule-based signals answer 'has this borrower been stressed' — synthetic-financial signals answer 'what business does this borrower run and how is it performing'. Unsecured MSME credit needs the second question answered.

Account Aggregator payloads arrive structured (transaction-level JSON with categorisation tags from the FIU), with verified provenance, and with consent-bounded scope. They are higher-trust than PDF-derived signals because they cannot be tampered with mid-flight. But AA payloads still need a downstream analytics engine because the FIP-emitted categorisation is coarse (debit/credit type at best) and does not produce the 40+ engineered signals lenders need for underwriting. AA solves authentication and ingestion; it does not solve analytics. Lenders running AA-only pipelines still need a BSA layer on top to derive bounce ratios, salary consistency, synthetic financials, and round-tripping flags from the raw transactions.

Depends on the vendor architecture. Closed-box vendors expose only the published signal list and do not let you add policy-specific derivations. Configurable vendors expose a rules layer where you can compose new signals from primitive transactions and existing engineered signals — for example, 'inflow from notified anchor employers as % of total inflow' or 'EMI obligation to GST turnover ratio' specific to your underwriting model. TransactIQ exposes derived signals at the transaction-classification layer and lets credit policy owners compose lender-specific signals on top of the 40+ engineered base. Discuss specifics during a scoping conversation; the marketing site does not publish the rule grammar.