IRAC Provisioning Platform
A complete Income Recognition, Asset Classification & Provisioning engine for Urban Co-operative Banks and NBFCs — built directly against the current RBI Directions, not a spreadsheet approximation of them.
Provisioning Modules
DPD-based asset classification (Standard / Sub-Standard / Doubtful / Loss)
Product-category LGD matrix, including CRE-RH as its own sub-category
Secured / unsecured exposure splitting per account
Project-finance construction-phase rate table
Board-configurable Provision Coverage Ratio overlay
One-click board-ready PPTX + CSV reporting
Full audit trail and role-based access control
Rates Sourced Directly From the Directions
Every provisioning rate in the platform is cited to a specific paragraph of the RBI (Urban Co-operative Banks – Income Recognition, Asset Classification and Provisioning) Directions, 2025 — not a commercial-bank schedule adapted by assumption, and not a decimal-point guess. When RBI revises a rate, the change is a config update, not a re-audit of the formula.
Asset Classification & Ageing
Every account is classified from its current, continuous days-past-due position — not its historical worst DPD — so a borrower who has genuinely cured is never resurrected into the current provisioning run together with a stale historical balance.
Standard
0 – 90 DPD. Provisioned at a baseline rate by product category (0.25% Agri/SME, 1.00% CRE, 0.75% CRE-RH, 0.40% all others) — no tier-based differentiation, per the current Directions.
Sub-Standard (NPA)
91 – 455 DPD — remained NPA for ≤ 12 months. Provisioned at a flat 10% of total outstanding, with no secured/unsecured split.
Doubtful
> 455 DPD — NPA for more than 12 months, banded by time spent in the Doubtful category: 20% (≤ 1 year), 30% (1–3 years), 100% (> 3 years) of the secured portion, and 100% of whatever isn't covered by realisable security value.
Loss
Identified as a loss by the bank, its auditors, or an RBI inspection — not derived from DPD. Provisioned at 100%, or written off entirely.
Why the split matters: a general provision applies uniformly to the whole Sub-Standard bucket, while Doubtful accounts are provisioned according to how long they've specifically been Doubtful — so two accounts with the same DPD count can carry different rates if one has been restructured or re-aged along the way. The engine tracks that distinction rather than collapsing every NPA into a single rate.
Provision Coverage Ratio — Correctly Benchmarked
The current RBI Directions for UCBs don't prescribe a Provision Coverage Ratio requirement — it's an optional discipline a Board can choose to hold itself to above the statutory minimum. Where it gets applied is where most spreadsheets and other tools go wrong.
Not Provision ÷ Total Portfolio
Benchmarking coverage against the whole book folds a bank's healthy Standard-asset book — where provisioning is a fraction of a percent by design — into the ratio. A well-covered, low-NPA bank would always look under-provisioned on that basis, and a floor applied to it would demand provisioning a slice of performing loans that were never at risk.
NPA Provisions ÷ Gross NPA
The platform computes PCR — and any Board-set floor — against Gross NPA specifically. If coverage falls short, only the NPA book gets topped up to policy; Standard-book provisioning, already computed correctly above, is never touched by the floor.
Governance, Audit Trail & Reporting
A provisioning platform is only as trustworthy as its controls. Authentication, role separation, and a full audit log are first-class features, not an afterthought bolted on before an inspection.
Full Audit Trail
Every login, password change, and user-management action — logged with timestamp, actor, and IP address, filterable and exportable for examiners and statutory auditors.
Role-Based Access
Admin, Risk Analyst, and Viewer roles with page-level enforcement, salted PBKDF2 password hashing, and automatic session timeouts.
One-click board reporting: a five-slide PPTX deck — KPI summary, full provisioning table, portfolio composition, and charts — plus a companion CSV with every figure used, so numbers are always independently reconcilable. Every report generated is logged to a searchable Report History with re-download access.
See the Book, Not Just the Total
Exposure share by IRAC bucket, by secured/unsecured status, and by product line — computed directly from the same source data as the provisioning grid, so the dashboards and the numbers they explain never drift apart.
Accounts with no matched security record default conservatively to Unsecured — the higher-provisioning outcome — rather than being assumed benign, with the resulting match rate surfaced as a data-quality metric so coverage gaps are visible, not hidden inside an averaged figure.