6.17 Graduated lending — small first, observe, grow

The single most effective risk-mitigation pattern for thin-file lending is graduated lending: extend a small, short-tenure first loan; observe; step up based on demonstrated repayment behaviour.

This works because:

Internal performance data trumps any external signal. A borrower’s repayment behaviour on your loan tells you more than the cleanest GST or bank statement ever could.
Small first loans limit downside. A ₹3 lakh first loan that defaults is recoverable; a ₹30 lakh first loan that defaults is a meaningful loss.
Repeat borrowers’ default rates are dramatically lower than first-time defaults. Industry data suggests 1.5 – 2.5× lower for borrowers on their second loan with the same lender after clean first-loan repayment.
Operational cost amortises over the relationship. The thin-file diligence (FI, references) was already done at first loan; subsequent loans cost much less to underwrite.

The graduation ladder

Concrete example for a thin-file C/D-grade borrower:

Loan #	Ticket	Tenure	Rate	Conditions
1 (first)	`₹2 – 5 lakh`	`30 – 60 days`	`24 – 26%`	Full thin-file diligence; field + references; tight EWS monitoring
2 (clean L1)	`₹5 – 10 lakh`	`45 – 90 days`	`22 – 24%`	Light re-verification; refreshed AA / GST
3 (clean L1, L2)	`₹10 – 20 lakh`	`60 – 120 days`	`20 – 22%`	Standard refresh; periodic FI
4 (clean L1-3)	`₹20 – 35 lakh`	`90 – 180 days`	`18 – 21%`	Standard process; bureau refresh
5+ (sustained clean)	Standard ticket	Standard tenure	Standard rate	Standard treatment, but with internal score override

The thresholds and pricing are illustrative; the principle is the ladder.

Definitions of “clean” between rungs

No DPD over 30 days during the loan life.
Full repayment (not partial or settled) at maturity.
No requests for restructuring during the term.
At least 60 days of clean post-closure history before the next loan.
External signals haven’t deteriorated between loans (no GST suspension, no MCA strike-off, no major bureau-derived red flag).

A borrower failing any of these doesn’t graduate; the next loan is either at the same rung or one step back.

Step-up rules

Maximum step-up: ~2× of previous loan's ticket per graduation.
Tenure step-up: +30 days per graduation.
Pricing step-down: ~100 – 200 bps per graduation (rewards clean behaviour).
No step-up if: bureau score has dropped materially; GST volatility has spiked; bank-statement signals deteriorated; FI re-verification flagged.

Step-down (demote) rules

A borrower’s behaviour can demote them down the ladder:

30+ DPD during the loan term → next loan at same or lower ticket.
60+ DPD → next loan only after gap of 6 months clean; lower ticket.
NPA classification → no further lending until full recovery + cooling-off.
Restructuring requested → next loan after 12-month observation period at smaller ticket.

Why this works for the platform’s economics

Graduated lending is fee-revenue rich in a way single-large-loan thin-file is not:

5 small loans of ₹2 / 5 / 10 / 20 / 35 lakh over 18 months = ₹72 lakh cumulative disbursement to one borrower.
Standard single-large-loan would have been a ₹35 lakh ticket at maximum.
Processing fees compound across loans.
Cost-of-acquisition (the expensive first-loan diligence) amortises across 5 loans, not 1.
Net margin per ₹ of capital ends materially higher than single-large-loan thin-file.

How the platform enables it

Repeat-borrower workflow (3.A) is light-touch; pre-fills the application from prior data.
Internal scorecard (3.E) heavily weights prior repayment behaviour for repeat borrowers.
Limit step-up engine in admin (3.O) configures the ladder per product / segment.
Borrower portal surfaces “you’re eligible for a higher limit” prompts at the right moment.
Communication cadence — at loan closure, immediately surface next-loan offer to the borrower.

A common adversarial pattern

Some borrowers see the graduation pattern and try to exploit it — clean first loan, larger second loan, then default. Mitigations:

Step-up amount cap in absolute terms (<= ₹2× prior or <= ₹X lakh, whichever is lower) — limits damage.
External refresh between loans — re-pull bureau, GST, BSA; deterioration blocks step-up.
Field re-visit for material step-up.
Network analysis — borrower who suddenly applies at multiple lenders during your graduation is a flag.
Behavioural triggers — sudden change in transaction pattern, sudden new high-value purchase order, etc.

When graduated lending isn’t the right pattern

One-time capital need — borrower needs ₹30 lakh for a specific purpose; graduated lending starts at ₹3 lakh; offer is meaningless. Either find a co-lending partner that takes the borrower at standard process, or decline.
Time-bound opportunities — borrower needs to act now (e.g., bid on a contract); ladder doesn’t fit. Field-FI-heavy single underwriting at higher pricing.
Borrowers who can prove themselves through anchor data — graduated lending isn’t needed; anchor-backed underwriting can sanction larger first loan.

Operational discipline

Document the ladder as a board-approved policy.
Track cohort graduation rates — what % of first-loan borrowers go to second loan? Third? This is a key product metric.
Track delinquency by rung — bucket borrowers by their current rung; expected default rates differ.
Cohort vintage analytics — graduation cohort by acquisition year shows whether the wedge is improving over time.

Outcomes

A graduated-lending book at maturity typically shows:

Repeat-borrower share > 50% of disbursements (a key health metric).
Average loan ticket increase YoY as the cohort matures.
Credit cost on repeat loans < 1.5% vs 2.5 – 4% on first-loans.
Operating cost per disbursal materially lower (light-touch repeat workflow).
Cohort profitability breakeven by loan #2 and accelerating from #3 onward.

These outcomes are the strategic argument for the patience required at the start.

6.14 Thin-file underwriting.
6.15 Unconventional data sources.
3.A Borrower acquisition — repeat-borrower workflow.
3.O Admin configuration — limit / pricing ladder config.
16. Client diligence — overall diligence framework.