3.L Portfolio monitoring
Purpose
Section titled “Purpose”Watch the book continuously. Detect concentration, drift, deterioration, and concentration risks before they show up in default. Provide the credit and risk teams with the visibility to act.
In-scope features
Section titled “In-scope features”Vintage and cohort
Section titled “Vintage and cohort”- Vintage analysis — DPD curves by month of disbursement; identify cohort-level deterioration.
- Cohort tagging — by product, channel, partner, geography, DSA, underwriter, policy version.
- Annualised DPD curves —
30+,60+,90+,180+over time.
Bucket analysis
Section titled “Bucket analysis”- DPD distribution — Standard, SMA-0, SMA-1, SMA-2, NPA, sub-buckets.
- Bucket-roll-rate matrix — monthly transitions between buckets.
- Cure rate — % of SMA upgrading back to Standard.
Cuts and dimensions
Section titled “Cuts and dimensions”- Per product — line vs term vs invoice-backed.
- Per partner for co-lent.
- Per industry — NIC code groupings.
- Per geography — state, district.
- Per DSA / channel.
- Per underwriter / decision-engine policy version.
- Per ticket bucket —
5L – 10L,10L – 25L,25L – 50L,50L+.
Performance metrics
Section titled “Performance metrics”- Disbursement — by period, by cohort.
- AUM / book — daily.
- Collection efficiency — % collected vs scheduled.
- Prepayment rate.
- Repeat-borrower rate.
Yield and economics
Section titled “Yield and economics”- Yield — actual yield realised vs IRR sanctioned.
- NIM (Net Interest Margin).
- Spread — yield minus cost of funds.
- IRR — per loan, per cohort.
- Risk-adjusted return — net of expected credit loss.
ECL and risk
Section titled “ECL and risk”- Expected Credit Loss (ECL) — per loan and aggregated; per IndAS 109 stage-1 / 2 / 3 (for IndAS-applicable entities).
- Provisioning mirrors classification but ECL extends — forward-looking.
- Stress testing — what-if scenarios (e.g., GST shock, sector shock, anchor default).
- Early warning signals — see below.
Early warning signals (EWS)
Section titled “Early warning signals (EWS)”Monitor signals that precede default:
- Bank-balance drop — borrower’s MAB falls below threshold (requires periodic AA re-pull).
- GST sales drop — month-on-month drop above threshold.
- GST late filings — late filing in successive periods.
- GSTIN status change — suspension, cancellation.
- Bureau score drop — material decrement vs onboarding score.
- NACH bounce uptick — pattern across recent attempts.
- Anchor relationship change — for SCF borrowers.
- Litigation hit — court case filing.
- Director resignation, change in shareholding — MCA pulls.
Each EWS surfaces an alert; the risk team triages.
Limit and concentration monitoring
Section titled “Limit and concentration monitoring”- Single borrower exposure vs Tier-1 capital.
- Group exposure.
- Sector exposure.
- Geography exposure.
- Anchor exposure.
- Partner-lender exposure (for own-book in partnership ecosystems).
- Underwriter / channel exposure caps.
Reports + dashboards
Section titled “Reports + dashboards”- Daily risk dashboard (CRO).
- Monthly board pack — book composition, vintage, NPA, ECL, key flags.
- Quarterly partner review packs.
- Drill-down to individual loans from any aggregate.
Out of scope
Section titled “Out of scope”- Trade journals / market data integration (out of scope — internal monitoring only).
- Recommendation of policy changes — manual; the engine produces evidence, humans decide.
Key entities
Section titled “Key entities”Cohort— definition + members.VintageBucket— DPD aging by cohort.Ews— alert with type, severity, loan reference, evidence.Limit— concentration / sectoral / channel cap.LimitBreach— when threshold crossed.
Key workflows
Section titled “Key workflows”- Daily DPD computation — feeds vintage and bucket.
- Weekly cohort refresh — vintage curves rebuilt.
- EWS computation — periodic (daily / on-data-arrival).
- EWS triage — risk team queue.
- Limit monitoring — real-time at sanction; daily aggregate.
- Periodic stress test — quarterly minimum.
Integrations
Section titled “Integrations”- Read from LMS, data-ingestion, KYC, bureau, AA, GST.
- Write to BI tools — Metabase, Superset, internal warehouse.
- Alert routing — Slack / email / on-call paging for EWS.
GET /portfolio/cohorts— cohort list.GET /portfolio/vintage/{cohort_id}— vintage curve.GET /portfolio/roll-rate?period=month— roll-rate matrix.GET /portfolio/ews— current EWS alerts.POST /portfolio/ews/{id}/dispose— triage outcome.GET /portfolio/limits— current exposure vs caps.
Events emitted
Section titled “Events emitted”ews.raised(high priority)limit.breach.warning/limit.breach.violationcohort.refreshedstress_test.completed
Edge cases
Section titled “Edge cases”- Cohort cuts too granular — empty buckets; minimum population threshold.
- Late data arrival — re-compute affected dates.
- EWS storm — circuit-breaker; batch and prioritise.
- Partner partial data — co-lent loans where partner data lags; reconcile before publishing partner-cut MIS.
- GST data delayed by 1 month (filing cycle) — EWS based on filing recency must account.
Compliance touchpoints
Section titled “Compliance touchpoints”- SBR concentration limits — monitored.
- IRACP — classification feeds.
- Bureau reporting — performance data feeds.
- Board governance — risk-committee reporting cadence.
MVP vs production
Section titled “MVP vs production”| Feature | MVP | Production |
|---|---|---|
| Vintage / cohort | Basic (monthly cohorts) | Full cuts |
| Bucket roll-rate | ✓ | ✓ |
| ECL | Simple (rule-based) | IndAS-aligned model |
| EWS | 4–6 signals | 15+ signals |
| Stress test | (Phase 3) | Quarterly automated |
| Limit monitoring | ✓ | ✓ |
| Partner cuts | ✓ (when co-lending live) | ✓ |
| Board dashboards | Manual MIS | Automated |
Related: 2.10 Asset classification, 11. Risk register, 3.P Analytics, 13.19 Analytics backlog.