Co-Lending Eligibility Filter: Pick Partner-Ready Loans
Last reviewed: November 2026
A co-lending eligibility filter is a rules engine that checks each loan against the partner bank's agreed selection criteria and tags it for co-lending only if it qualifies. When that screening is manual, ineligible loans reach the partner, are disputed after disbursement and stay on the NBFC's balance sheet at full exposure.
This guide is for heads of co-lending and CROs at NBFCs, housing finance companies and banks in co-lending arrangements (CLAs). It covers what RBI's co-lending rules mean for loan selection, why loans fail partner criteria, a seven-step filtering workflow, what late rejection costs and the metrics to track.
Key Takeaways
- RBI's co-lending rules make the partner's commitment irrevocable and back-to-back, so eligibility must be settled before disbursement.
- A loan whose partner share is not booked within 15 calendar days stays with the originating NBFC.
- Each lender must keep at least 10 per cent of every loan, and the agreement must state the borrower selection criteria.
- Versioned rules, stamped on each loan, end most arguments about which criteria applied.
What Is a Co-Lending Eligibility Filter?
A co-lending eligibility filter turns the partner's written selection criteria into system rules that run on every loan. It returns one of three answers: eligible for a named partner, eligible with an exception to review, or not eligible.
Three terms matter. The originating RE sources and disburses the loan, the partner RE co-funds an agreed share, and back-to-back means the partner takes its share on the same terms as the original loan.
What Do RBI's Co-Lending Rules Say About Loan Selection?
RBI's co-lending rules shape loan selection in five ways. RBI issued the Co-Lending Arrangements Directions, 2025 on 6 August 2025, effective 1 January 2026, and carried them into Part B of the NBFC Transfer and Distribution of Credit Risk Directions, 2025 (28 November 2025).
- Ex-ante agreement (para 101): co-lending is formalised through an agreement made in advance.
- Retention (para 102): each lender keeps a minimum 10 per cent share of each loan.
- Selection criteria (para 104): the agreement must include the criteria for selection of borrowers, products and geographies.
- Irrevocable commitment (para 112): the partner commits to take its share of each loan back-to-back.
- 15-day window (paras 113 and 115): shares must be booked within 15 calendar days of disbursement, or the loan stays with the originator.
Why Do Loans Fail the Partner Bank's Criteria?
Loans fail the partner bank's criteria for six common reasons: criteria held in documents, unversioned changes, different data definitions, stale bureau data, missed geography or concentration limits and unflagged policy deviations. Each has a prevention check.
| Cause | Effect | Prevention Check |
|---|---|---|
| Criteria kept in a PDF or email | Credit teams apply them from memory | Codified rules in the loan system |
| Criteria changed without versioning | Dispute over which rules applied | Rule version stamped on every loan |
| Income or obligation defined differently | Loan passes your check, fails theirs | Field mapping agreed with partner |
| Bureau report older than partner allows | Rejection on data freshness | Bureau age check at disbursement |
| Pincode or concentration limit missed | Rejection or partner limit breach | Geography and limit checks in the filter |
| Credit policy deviation not flagged | Exception loan sent as standard | Deviation flag excludes or routes for review |
How Do Policy Deviations Affect Eligibility?
Policy deviations affect eligibility because most partners accept only loans approved within policy, or a capped share of exceptions. Our guide to credit policy deviation tracking explains how to record deviations so the filter can act on them.
What Is the Right Co-Lending Eligibility Filter Workflow?
The right co-lending eligibility filter workflow has seven steps: codify the criteria, map the data, run the filter before sanction, allocate to a partner, lock eligibility at disbursement, book within 15 days and learn from rejections. Each step moves a decision earlier, where it is cheaper.
- Codify the criteria as versioned rules with effective dates.
- Map the data so each rule reads the field the partner means.
- Run the filter at sanction, before the borrower is promised terms.
- Allocate to a partner using limits, pricing and concentration.
- Lock eligibility with a final check at disbursement.
- Book within 15 days and confirm the partner's share.
- Learn from rejections and update rules with the partner.
Step 1: How Should Partner Criteria Be Codified?
Codify partner criteria as rules, each with a data source, a threshold and an effective date. Every loan should carry the rule version it passed, so both lenders can see which criteria applied.
| Criterion Type | Example Rule | Data Source |
|---|---|---|
| Product and purpose | Only listed products and end uses | Loan application |
| Ticket size and tenure | Within partner's minimum and maximum | Sanction terms |
| Bureau | Score and recent delinquency within limits | Credit bureau report |
| Income and obligations | Ratio within partner's threshold | Underwriting data |
| Geography | Pincode on partner's allowed list | Address and pincode |
| Documentation | All mandatory documents verified | Document checklist |
Step 2: Why Does Data Mapping Matter?
Data mapping matters because two lenders can use the same word for different numbers. Agree with the partner how income, obligations and vintage are calculated, then map each rule to that definition.
Clean inputs also matter. Loan application deduplication stops the same borrower being sent twice, and a maintained negative pincode list keeps geography rules current.
Step 3: When Should the Filter Run?
The filter should run at sanction, before terms are shared with the borrower, because the blended rate depends on whether the loan is co-lent. Running it only after disbursement turns every failure into a balance sheet problem.
Our guide to AI agents in loan underwriting covers how underwriting data can be prepared so these checks run automatically.
Step 4: How Should Loans Be Allocated Among Partners?
Allocate eligible loans using agreed priorities: partner limits, pricing, concentration caps and any committed volumes. Digiqt's loan participation allocation AI agent shows how loans can be allocated across funding partners by rule.
Loans that fit no partner stay on your own book by design. That decision should be visible to treasury on the day of sanction.
Steps 5 and 6: What Should Happen at Disbursement and Booking?
At disbursement, re-run the rules that depend on changing data, such as bureau age, documents and partner limits. Lock the result and send the partner's share for booking with a loan-level data file.
A consistent file format, similar to a securitisation data tape, speeds the partner's booking. Track every loan against the 15-day window, because loans not booked in time stay with the originator.
Step 7: How Should Rejections Improve the Filter?
Rejections should feed straight back into the rules. Review the partner's queries and rejections monthly, rank them by rule, and fix the top causes with the partner.
How many loans tagged for co-lending last quarter ended up staying on your own book?
Talk to Our Lending Platform Specialists
Visit Digiqt to see how a co-lending eligibility engine can screen every loan before it is disbursed.
How Much Do Ineligible Co-Lending Loans Cost?
Ineligible co-lending loans cost the originator extra funding and capital, because the whole loan stays on its book. In this illustrative scenario, an NBFC originates ₹200 crore a month for an 80:20 arrangement and 6% of it, ₹12 crore, is found ineligible after disbursement.
The NBFC planned to hold 20%, or ₹2.4 crore, but now holds all ₹12 crore, so it must fund an extra ₹9.6 crore. At a 10% cost of funds, that is ₹8 lakh a month in interest for each month's cohort.
| Measure | 6% Found Ineligible | 1% Found Ineligible |
|---|---|---|
| Ineligible loans a month | ₹12 crore | ₹2 crore |
| Extra on-book exposure (80% share) | ₹9.6 crore | ₹1.6 crore |
| Extra funding cost a month at 10% | ₹8,00,000 | ₹1,33,333 |
| Extra capital at an assumed 15% | ₹1.44 crore | ₹24 lakh |
The lesson for leadership: the cost compounds, because each month's ineligible loans stay for their full tenure. Moving the check before disbursement is cheaper than any amount of post-facto negotiation.
Where Does the Digiqt Lending Platform Fit in Co-Lending Loan Selection?
The Digiqt Lending Platform is a custom lending product that covers onboarding, disbursement and repayments, delivered through APIs. Digiqt says such a platform can be launched in a matter of weeks.
The page does not list a ready-made co-lending feature, so a lender would specify an eligibility engine as part of a custom build. Such an engine can be built to include:
- Versioned partner rules with effective dates, stamped on every loan.
- Field mapping to each partner's data definitions.
- Multi-partner allocation by limits, pricing and concentration.
- A 15-day booking tracker and rejection analysis by rule.
Digiqt's process runs from problem understanding and solution design to agile delivery, launch and maintenance. Our guides to AI agents in co-lending and voice agents in co-lending cover the borrower communication side.
Which Co-Lending Eligibility Filter Design Fits Your Setup?
The right design depends on how many partners you have and how often their criteria change. Use this table as a starting point.
| Co-Lending Situation | Recommended Approach |
|---|---|
| One partner, one product | Rules in the loan system with version stamps |
| Several partners, overlapping criteria | Central eligibility engine with allocation rules |
| Partner criteria change often | Rule management screen with maker-checker approval |
| High exception volume | Separate exception route with partner pre-approval |
| New partner onboarding | Back-test rules on last quarter's loans before go-live |
Which Metrics Show the Co-Lending Eligibility Filter Is Working?
Five metrics show whether a co-lending eligibility filter is working: eligibility hit rate, partner rejection rate, 15-day booking rate, partner limit utilisation and rule version mismatches. Review them monthly with the partner.
| Metric | Formula | What a Weak Number Signals |
|---|---|---|
| Eligibility hit rate | Loans tagged eligible ÷ loans sanctioned in co-lent products | Sourcing outside partner appetite |
| Partner rejection rate | Loans rejected or disputed by partner ÷ loans submitted | Filter and partner criteria drifting |
| 15-day booking rate | Partner shares booked within 15 days ÷ loans submitted | Slow data exchange or disputes |
| Partner limit utilisation | Partner exposure ÷ partner limit | Over- or under-allocation |
| Rule version mismatches | Loans checked on an outdated rule version ÷ loans checked | Weak rule change control |
Which Mistakes Send Ineligible Loans to Co-Lending Partners?
Four mistakes send ineligible loans to partners, and all of them are design choices.
- Screening after disbursement. Every failure then lands on your balance sheet.
- Keeping criteria in documents. People apply what they remember, not what was agreed.
- Changing rules without versions. Disputes become arguments about which PDF applied.
- Ignoring rejection patterns. The same rule fails every month until someone fixes it.
Want every loan screened against your partner bank's criteria before disbursement?
Talk to Our Lending Platform Specialists
Visit Digiqt to see how the Digiqt Lending Platform can be built around your co-lending selection rules.



