A credit exception that is genuinely exceptional is a sign of good credit judgement — the policy handles 95% of cases and the credit team exercises sound discretion on the other 5%. A credit exception that is repeated, concentrated, and predictable is something different: it is evidence that the policy parameter causing the exceptions is miscalibrated for the market the institution is serving, or that a subset of the institution's origination sources is systematically producing borrowers who fall outside the policy for reasons that are not random. The Credit Exception Agent AI analyses exception data quarterly and surfaces these patterns — because a policy that generates systematic exceptions is a policy that needs updating, not a credit team that needs better discipline.
A credit exception that is genuinely exceptional is a sign of good credit judgement — the policy handles 95% of cases and the credit team exercises sound discretion on the other 5%. A credit exception that is repeated, concentrated, and predictable is something different: it is evidence that the policy parameter causing the exceptions is miscalibrated for the market the institution is serving, or that a subset of the institution's origination sources is systematically producing borrowers who fall outside the policy for reasons that are not random. The Credit Exception Agent AI analyses exception data quarterly and surfaces these patterns — because a policy that generates systematic exceptions is a policy that needs updating, not a credit team that needs better discipline.
The three types of exception pattern — and what each signals
The first pattern is parameter concentration: a single policy parameter — most commonly CIBIL floor or FOIR ceiling — accounts for a disproportionate share of exceptions. If 68% of all exceptions in Q3 were CIBIL score exceptions, and the institution is approving most of them because GST data and bank statement analysis indicate creditworthiness that the CIBIL score does not yet reflect, then the CIBIL floor is calibrated for a borrower population that does not describe the institution's Tier 2/3 MSME target market. The policy is right for a metro salaried borrower and wrong for a rural proprietor. This is not a credit team problem — it is a policy design problem that requires a policy revision, not a training programme.
The second pattern is source concentration: exceptions are clustering around borrowers sourced from specific DSAs, geographies, or channels. If 44% of all exceptions are sourced from 3 DSAs who together account for 18% of origination volume, those DSAs are either specialising in borrowers who fall outside the policy but are creditworthy (which is a market opportunity the policy is blocking) or they are systematically presenting edge-case borrowers for exception approval because they have learned that the approval process is permissive (which is a governance problem). The pattern looks identical before the DPD analysis; the DPD data distinguishes between the two.
The third pattern is officer concentration: a disproportionate share of exceptions are being initiated by a small number of relationship managers or credit officers. This can reflect positively (some RMs work in markets that genuinely generate more complex credit profiles) or negatively (some RMs have learned that exceptions are approved if presented correctly and are using the exception process as a workaround for rigorous credit assessment). Again, DPD data on the exception cohort associated with each RM disambiguates.
"A policy that generates 40% exceptions on CIBIL floor is not being enforced — it is generating paperwork. The Credit Exception AI asks: should the floor be recalibrated, or are 40% of our MSME borrowers genuinely exception cases?"
The quarterly exception pattern analysis: Q3 FY26
Exception Pattern Analysis — Q3 FY26 (Oct–Dec 2025) · 847 Exceptions · ₹124 Cr Total
Analysis date: Jan 3, 2026 · Pattern 1 of 4 triggers Board-level policy review · Pattern 3 triggers governance investigation
847Total exceptions Q3
4.8%Exception portfolio rate (vs 8% limit)
6.1%Exception 90-day DPD rate
3.8%Standard portfolio 90-day DPD rate
01
Parameter concentration — policy calibration signal
68.4% of all exceptions are CIBIL floor breaches — but the CIBIL breach cohort has a 90-day DPD of only 5.2%
580 of 847 exceptions were approved on CIBIL scores below the 680 minimum. Of these, 88.4% were MSME borrowers where the compensating factor was GST revenue growth. The CIBIL breach exception cohort has a 90-day DPD of 5.2% — 1.4pp above standard portfolio but well within the 3pp tolerance the Board has set for exception cohorts. The volume of these exceptions (580 in one quarter) suggests the 680 floor is miscalibrated for MSME borrowers in Tier 2/3 markets.
→ Signal: CIBIL floor of 680 may be 20–30 points too high for MSME Tier 2/3 segment · Exception frequency is structural, not random
Recommended action: Board credit policy review — consider a differentiated CIBIL floor by borrower segment: 680 for salaried/metro, 640–650 for MSME Tier 2/3 with GST vintage >24 months. If adopted, 580 quarterly exceptions reduce to an estimated 80–120, freeing credit team capacity for genuinely complex cases.
02
Source concentration — DSA channel investigation triggered
DSA group "Vijay Associates" (3 DSAs, same principal) sourced 112 exceptions — 13.2% of all exceptions from 6.8% of origination volume
DSAs VS-014, VS-022, and VS-031 — all operating under a single principal — sourced 112 exceptions this quarter versus 58 in Q2. Their exception rate (ratio of exception approvals to total applications) is 28.4% versus the DSA network average of 9.1%. Their exception cohort 90-day DPD is 9.8% — significantly higher than both the standard portfolio (3.8%) and the broader exception cohort (6.1%). This DSA group's exception borrowers are performing 2.6× worse than the standard portfolio.
→ Signal: DSA group is systematically presenting exception applications, and the performance data indicates the exceptions are not well-compensated · Governance concern
Recommended action: Immediate investigation — DSA principal meeting within 5 working days. Review of all 112 Q3 exceptions for documentation quality. If DPD persists at Q4 review: commission hold, exception pre-approval requirement, and potential DSA deregistration. Credit team approval process for these DSAs tightened immediately.
03
Geography concentration — market opportunity identified
Rajkot district: 84 FOIR exceptions, all MSME, all performing — policy FOIR ceiling may be wrong for this market
Rajkot has generated 84 FOIR exceptions this quarter — all MSME, all above the 65% ceiling, most in the 68–72% range. The Rajkot exception cohort 90-day DPD is 2.1% — lower than the standard MSME portfolio (3.4%) and significantly below the overall exception cohort (6.1%). These are the best-performing exceptions in the portfolio. Rajkot's MSME income profile is characterised by high seasonal income concentration (diamond cutting businesses with Q4-heavy cash flows) that produces higher FOIR during non-peak months without indicating actual over-leverage.
→ Signal: FOIR ceiling of 65% is wrong for Rajkot's seasonal income profile · Exception approvals are outperforming standard portfolio
Recommended action: Consider a geography-specific FOIR ceiling of 72% for Rajkot MSME borrowers with verified seasonal income documentation. If adopted, 84 quarterly exceptions reduce to near-zero while preserving a market the institution is serving well.
04
Positive pattern — exception quality improving
Q3 exception DPD (6.1%) has fallen from Q1 (9.4%) — exception quality improving as documentation and compensating factor standards tighten
The 90-day DPD rate of the exception cohort has improved from 9.4% in Q1 FY26 to 7.2% in Q2 and 6.1% in Q3. The number of exceptions with all 9 documentation fields complete has risen from 62% in Q1 to 94% in Q3. The Credit Exception AI's documentation completeness requirement — which returns incomplete records for additional data before approver routing — is producing better-documented exceptions and, as a result, better credit decisions. Exceptions that are hard to document tend to be exceptions where the compensating factors are weak.
→ Signal: Documentation quality improvement is causing credit quality improvement · Harder to approve poorly-documented exceptions = fewer bad exceptions approved
Recommended action: Continue documentation completeness enforcement. Present Q3 trend to Board Risk Committee as evidence that the exception governance programme is improving exception credit quality, not just exception record quality.
68.4%CIBIL exceptions as share of all exceptions — parameter concentration signal · Policy floor may be wrong for MSME Tier 2/3 segment · Not a team discipline problem
9.8%Vijay Associates exception cohort DPD — 2.6× the standard portfolio · Source concentration + adverse performance = governance investigation
2.1%Rajkot exception cohort DPD — below standard MSME portfolio · FOIR ceiling wrong for seasonal income profile · Market opportunity, not risk
9.4→6.1%Exception DPD trend Q1→Q3 — documentation quality improvement is producing credit quality improvement · Presented to Board as governance progress
The same exception data that reads as a compliance problem is also a product strategy document — if the institution knows how to read it
The 84 Rajkot MSME FOIR exceptions performing at 2.1% DPD are not a compliance failure — they are evidence that the institution has found a market it can serve at a higher FOIR ceiling than its policy currently allows, and is serving it well. The institution that reads this exception data as a policy compliance problem and tightens the FOIR ceiling in Rajkot will lose that market to a competitor who understood the seasonal income profile and wrote a policy for it. The institution that reads it as a market signal will propose a geography-specific FOIR ceiling to the Board, adopt it, and eliminate 84 unnecessary quarterly exceptions while deepening its penetration in a market where its credit quality is excellent. The Credit Exception Agent AI's pattern analysis turns the exception register from a compliance document into a product and policy intelligence tool — and the information it contains is available for free, because the institution already has it.
