17.6 Recovery analytics
Recovery without analytics is reactive — fire-fighting one NPA at a time. Recovery analytics turns the recovery function into a continuously improving operation.
This page is the metric inventory and the analytics framework.
Cohort-recovery curves
Section titled “Cohort-recovery curves”The single most important recovery metric.
Definition
Section titled “Definition”For each NPA cohort (loans turning NPA in a defined month), track cumulative recovery rate over time:
NPA cohort = loans newly classified NPA in month M
Day 0: 0% recoveredDay 30: X%Day 60: Y%...Day 360: Z%Day 720: W%Each cohort produces a recovery curve. Curves compare across cohorts to see if recovery operation is improving.
Sample curves
Section titled “Sample curves”Cumulative recovery rate %100% ┤ 90% ┤ 80% ┤ ─── ← Cohort Jan 2024 (legacy approach) 70% ┤ ───────── 60% ┤ ──────────── 50% ┤ ──────────── ─── ← Cohort Jan 2025 (improved) 40% ┤ ────────── ─────────────── 30% ┤ ──── ──────────── 20% ┤─── ────────── 10% ┤ ────────── 0% └────────────────────────────────────────────────────────────────── 0 90 180 270 360 450 540 630 720 Days since NPA classificationIf Jan-2025 cohort outperforms Jan-2024 at the same age, your recovery operation has improved.
Drivers
Section titled “Drivers”- Cohort composition (segment / ticket / product).
- Recovery effort allocated (number of agents, escalation paths).
- Legal pipeline efficiency.
- Settlement policy generosity.
- Macroeconomic environment.
What to plot
Section titled “What to plot”- One curve per cohort month.
- Aggregate “year-cohort” curves.
- Per-segment cohort curves.
- Per-ticket-band cohort curves.
- Per-channel cohort curves.
Roll-rate matrices
Section titled “Roll-rate matrices”Bucket-to-bucket transition rates per month.
| From / To | Standard | SMA-0 | SMA-1 | SMA-2 | NPA |
|---|---|---|---|---|---|
| Standard | 95.2% | 4.5% | 0.2% | 0.1% | 0.0% |
| SMA-0 | 75% | 10% | 12% | 2% | 1% |
| SMA-1 | 50% | 15% | 15% | 15% | 5% |
| SMA-2 | 25% | 15% | 15% | 25% | 20% |
| NPA | 5% | 2% | 3% | 5% | 85% |
The matrix shows:
- % curing from each delinquent bucket.
- % rolling worse.
- % staying stuck.
Deteriorating matrix month-over-month is an early warning.
Recovery cost ratio
Section titled “Recovery cost ratio”Recovery cost / amount recovered- For SMA-0 / 1: very low (mostly automated reminders).
- For SMA-2: medium (calls, visits).
- For NPA: high (legal pipeline, agent intensity).
- For settled / written-off: high.
Total cost / total recovered = blended recovery cost ratio. Industry typical: 5 – 15% of recovered amount.
If your ratio climbs without proportional recovery improvement, the recovery function is becoming inefficient.
Agent / agency performance
Section titled “Agent / agency performance”Per agent / agency metrics:
- Cases assigned.
- Cases resolved (cure / settled / closed).
- Resolution rate = resolved / assigned.
- Average resolution time.
- Recovery
₹ per ₹of cost. - Compliance flags (call sampling QA breaches).
- Complaint count.
Disposition pattern
Section titled “Disposition pattern”Per agent, distribution of call dispositions:
PTPcount.RNR (Refused / No Response).Wrong number.Busy.Refused.
Outlier agents (very high RNR rate vs peers) flagged for QA + training.
Settlement-rate analysis
Section titled “Settlement-rate analysis”For settled loans, track:
- Average haircut % by bucket age.
- Settlement-to-original-outstanding distribution.
- Time from NPA to settlement.
- Borrower segment / ticket / channel drivers.
Insights:
- Are we over-settling (giving away too much)?
- Are we under-settling (chasing for years instead of closing)?
- Which borrower segments respond best to settlement offers?
Legal-pipeline ROI
Section titled “Legal-pipeline ROI”For each legal case (arbitration / civil / DRT / SARFAESI):
- Case opened date.
- Path chosen.
- Cost incurred (legal fees, court costs).
- Outcome (judgment / settlement during proceedings / pending / dismissed).
- Recovery from outcome.
- Net = recovery
-cost.
Per legal path, compute:
- Median time to resolution.
- Median recovery rate.
- Median net (recovery
-cost).
This informs path-selection policy.
Write-off cohort analysis
Section titled “Write-off cohort analysis”For loans written off:
- Pre-write-off characteristics — what made them un-recoverable?
- Post-write-off recovery — sometimes substantial (cash-basis recognition).
- Lifetime recovery (pre + post write-off) per cohort.
Pattern recognition:
- Specific segments / channels with high write-off rates → underwriting feedback.
- Specific recovery actions that improved post-write-off recovery → operational learning.
Early-warning effectiveness
Section titled “Early-warning effectiveness”For loans that turned NPA, retrospective:
- Was an EWS raised before NPA?
- Was the EWS triaged appropriately?
- Could earlier action have prevented NPA?
EWS effectiveness rate = % of NPAs that had a triggered EWS before classification. Higher is better; below 50% suggests EWS engine is weak.
Per-product recovery profile
Section titled “Per-product recovery profile”| Product | Avg recovery rate | Avg time to recovery | Recovery cost ratio |
|---|---|---|---|
| Revolving WC line | … | … | … |
| Term loan | … | … | … |
| Invoice-backed | … | … | … |
| Anchor-led SCF | … | … | … |
| Co-lent | … | … | … |
Cross-product comparison guides product mix and pricing decisions.
Per-channel recovery profile
Section titled “Per-channel recovery profile”| Channel | NPA rate | Recovery rate | Net contribution |
|---|---|---|---|
| Direct | … | … | … |
| DSA | … | … | … |
| CA | … | … | … |
| Anchor | … | … | … |
Channels with high NPA + low recovery are net-negative; consider reducing allocation.
Per-vintage analysis
Section titled “Per-vintage analysis”For loans disbursed X months ago, current default status:
0 – 6 monthspost-disbursement: typical “honeymoon” default rate.6 – 12 months: stabilising.12 – 24 months: full picture.
Cohort comparison shows whether underwriting has improved over time.
Stress-scenario projections
Section titled “Stress-scenario projections”Periodic stress tests:
- “If credit cost doubles, what’s the impact on AUM, CRAR, P&L?”
- “If 10% of book moves to SMA-2 next quarter, what’s recovery capacity?”
- “If a major anchor exits, what’s the contagion?”
Drives capital planning and recovery-team sizing.
What the platform must build
Section titled “What the platform must build”- Cohort-tracking ETL — every NPA loan tagged to its cohort month.
- Recovery-rate computation updated daily.
- Roll-rate matrix generated monthly.
- Per-agent / agency dashboards.
- Settlement / restructure / legal pipeline analytics.
- EWS effectiveness retrospective.
- Reporting cadence — daily dashboards for operations, weekly for risk team, monthly for CRO + board.
Operating cadence
Section titled “Operating cadence”- Daily: ops dashboard view.
- Weekly: recovery team review meeting.
- Monthly: CRO + recovery head review with action plan.
- Quarterly: board risk committee review.
Board-level metrics
Section titled “Board-level metrics”For the board’s risk committee:
- Cumulative NPA % of AUM.
- Recovery rate on NPA pool YTD.
- Provisioning vs realised loss.
- Cohort-curve trends.
- Recovery cost ratio.
- Settlement / restructure / legal pipeline summary.
- Write-off pattern.
- EWS effectiveness.
These metrics, tracked over time, give the board the picture of how recovery is performing.
What good looks like
Section titled “What good looks like”Mature recovery operation shows:
- Cohort recovery curves steepening over time (newer cohorts recover faster).
- Roll-rate matrix improving (fewer roll-forward, more cure).
- Recovery cost ratio stable or declining.
- Settlement / restructure / legal mix balanced (no over-reliance on any one tool).
- EWS effectiveness rising.
- Write-off rate predictable and declining.
Compliance touchpoints
Section titled “Compliance touchpoints”- DPDP — analytics on borrower data must respect purpose limitation.
- Board governance — recovery analytics is a board-reportable area.
- Internal audit — periodic audit of recovery analytics methodology.
Related
Section titled “Related”- 3.L Portfolio monitoring — broader portfolio analytics.
- 3.P Analytics and intelligence — analytics module.
- 17.1 Strategy by bucket.
- 11. Risk register.