Consumer Finance Portfolio and Receivable Risk Review for an NBFC

Where is the loss being manufactured, and is it a credit problem, a process problem or a policy problem?

Retail books fail differently from corporate books. In corporate lending, loss concentrates: one exposure deteriorates and the institution knows which one. In retail, loss is manufactured at the level of a rule. A relaxed underwriting criterion, a collection process that stops chasing after a fixed number of attempts, a channel with weak verification, a geography with poor recovery infrastructure, each of these produces loss across thousands of small accounts simultaneously. No single account tells the institution anything. The population does. The review therefore treated the book as a statistical object first and as a set of credit decisions second.

Five workstreams carried the review, moving from the shape of the book to the mechanics that produced that shape.

The portfolio was segmented and analysed by vintage so that the performance of accounts originated in each period could be compared at the same age, the only reliable way to separate origination-driven deterioration from external conditions. Receivables were analysed by ageing bucket with roll rate behaviour assessed between buckets, revealing what static ageing conceals: whether accounts are entering delinquency faster or leaving it slower. The receivable risk framework was reviewed as designed and as operated, because the gap between the two is where most retail loss originates. Collection processes were tested against policy: contact discipline, escalation triggers, settlement authority, field controls and legal handover. Provisioning was assessed against the behaviour actually observed in the current book rather than against the historical assumptions the model was built on.

Three disciplines separated this engagement from portfolio reviews that locate loss but leave management to infer the cause.

First, data validated before analysed. High-volume retail data is uneven, and analysis on unvalidated extracts produces conclusions that are confident and wrong. The data quality exceptions were themselves a finding.

Second, cause classified, not just loss located. Identifying where loss sits is straightforward. Establishing whether it originates in credit, process or policy is what determines the remedy, and it is the distinction most portfolio reviews leave to management to infer.

Third, designed against operated. The framework was tested as it actually runs under volume pressure, not as documented. Controls that exist in policy and are bypassed in practice under volume pressure produce exactly the loss pattern the policy was written to prevent.