Interactive demo · synthetic data · signals, not decisions
Co-lending controls, recomputed from both lenders’ records
Choose a scenario. The generator builds a small two-lender arrangement with that defect injected, and the same control engine the product runs recomputes every control from the synthetic records. Nothing below is hand-written — each finding is the engine’s own verdict on the data it was given.
Clean arrangement
Nothing. Both lenders' records agree and every mandatory field is present.
Retention floor breached
One loan where the originating lender keeps less than the 10% minimum share.
Booking window missed
A loan whose partner booking falls outside the 15-calendar-day window.
Lenders classify the same loan differently
The originator carries the borrower as SMA while the partner still shows STANDARD.
Duplicate escrow credit
The same collection is credited twice in the escrow feed.
Mandatory data absent
Fields the controls depend on are missing from the loan tape.
Default-loss cover (DLG) over the cap
A default-loss cover (DLG) arrangement above the permitted ceiling.
Only one lender reported to the bureau
The partner's share of the loan never reaches the credit bureau.
Fee outside the key-facts statement
A charge the borrower pays that the key-facts statement does not disclose.
Loan exists in one book only
A loan with an account in the originator's books and none in the partner's.
Gap in the escrow statement series
A missing statement sequence number in the escrow bank feed.
Priority-sector target not met
The bank's co-lent book falls short of its aggregate PSL target.
Scenario
Booking window missed
- Originator
- Kosha Micro
- Partner
- Sahyadri Bank
- Loans in book
- 3
- Evaluated as of
- 2026-09-05
- Open findings
- 3
- Rule pack
- co-lending-pack-1 v1
What was injected: A loan whose partner booking falls outside the 15-calendar-day window.
What should surface: CL-TRF-01 fails; RED.
Indicators
- Compliance CoverageRED
- Operational HealthUNKNOWN
- Data QualityGREEN
- Financial ReconciliationGREEN
- Governance ReadinessUNKNOWN
An indicator that reads UNKNOWN is telling you the engine could not evaluate — it is never rendered as compliant.
manifest 70f217b2f2396222…
Findings (3)
- CL-TRF-01FAILHIGH · non-maskableloan cla-50-loan-1
transfer window expired (25d > 15d). Loan stays wholly on the originating RE's books; may move only under MD-TLE 2021 — not force-fed to the partner. Exits CLA treatment. (RBI Co-Lending Directions 2025, Para 24.)
- CL-TRF-01FAILHIGH · non-maskableloan cla-50-loan-2
transfer window expired (24d > 15d). Loan stays wholly on the originating RE's books; may move only under MD-TLE 2021 — not force-fed to the partner. Exits CLA treatment. (RBI Co-Lending Directions 2025, Para 24.)
- CL-TRF-01FAILHIGH · non-maskableloan cla-50-loan-3
transfer window expired (23d > 15d). Loan stays wholly on the originating RE's books; may move only under MD-TLE 2021 — not force-fed to the partner. Exits CLA treatment. (RBI Co-Lending Directions 2025, Para 24.)
The 24 controls, with this run’s result
A control raises a finding only when it fails or cannot evaluate. “No finding” means it either passed or did not apply to this arrangement — the product’s run record distinguishes the two; this demo shows the findings.
| Control | Source | Result | Note |
|---|---|---|---|
| CL-RET-01 | Para 10 | no finding | Each RE must retain ≥10% of each individual loan on its own books (per-loan, per-RE floor; no RE books >90% of any loan). 2025 replaced the old 20/80 split with a symmetric 10% floor. |
| CL-TRF-01 | Para 22, 24 | FAIL | Partner share must be reflected in both REs' books within 15 calendar days of the originating RE's disbursement; on expiry the loan stays wholly with the originating RE (MD-TLE only). |
| CL-RATE-01 | Para 17-19 | no finding | Borrower charged one blended rate = weighted average of each RE's rate by proportionate funding share. |
| CL-ESC-01 | Para 26 | no finding | All disbursements and repayments route through an escrow account with a bank. |
| CL-ESC-02 | Para 26 | no finding | Escrow credits reconcile against themselves (unique by end-to-end id): duplicate, partial and unmatched lines are findings. Self-consistency only — a missing credit is not detectable from this feed alone. |
| CL-CLASS-01 | Para 33 | no finding | Borrower-level classification: if one RE is SMA/NPA on CLA default, the other must match; divergence is a breach. |
| CL-DQ-01 | DQ | no finding | Mandatory inputs must be present to evaluate the pack; missing data yields UNKNOWN, never PASS. |
| CL-DLG-01 | Para 32 | no finding | Originating RE DLG ≤ 5% of loans OUTSTANDING under the CLA (base = outstanding, authoritative). |
| CL-DLG-02 | Para 20, Footnote 3 | no finding | Fees must not embed any direct/indirect credit enhancement or implicit guarantee of portfolio performance. |
| CL-PSL-04 | Paras 9-16 | no finding | A loan claimed as PSL must fall within its category ceiling (each RE claims its own share, Co-Lending Para 15). |
| CL-CLASS-02 | Para 31 | no finding | Each RE must report its own share to the credit information companies (CICRA 2005) — a single filer is a breach. Divergent DPD/status between the two filings, or an unacknowledged/rejected file, is a reportable exception. |
| CL-TRF-02 | Para 24, 23 | no finding | If the 15-day window expires, the loan remains wholly on the originating RE's books and may thereafter move only under MD-TLE 2021 — it must NOT be force-fed to the partner under CLA treatment. Any transfer may only be to the ex-ante-named partner. |
| CL-RATE-02 | Para 19, 14, 36 | no finding | Every fee/charge beyond the blended rate must be folded into the APR and disclosed in the Key Facts Statement (RBI KFS circular, 15 Apr 2024). This is a conduct control: it tests what the BORROWER was actually shown, not merely what the lenders computed. |
| CL-PSL-01 | Para 15 | no finding | PSL status is claimable per-RE for that RE's own share of the credit — never the whole loan by either party. Double-claiming the full amount across both REs would inflate PSL achievement. |
| CL-PSL-02 | Para 5, Para 30(vi) | no finding | A PSL loan sits in exactly ONE of the eight categories (Agriculture, MSME, Export Credit, Education, Housing, Social Infrastructure, Renewable Energy, Others). Zero categories is unclassifiable; more than one is double-counting. |
| CL-PSL-05 | Para 17.1 | no finding | Weaker-sections eligibility list (Para 17.1). Machine-testable slice: individual women beneficiaries count toward weaker sections only up to ₹2 L per borrower — the cap STANDS (it does not apply to UCBs). List-membership beyond that is attester judgement, not engine. |
| CL-PSL-06 | Para 1.2 / 4.3 / 26 | no finding | Loans PSL-eligible under the 2020 MD remain eligible till maturity (grandfathering); co-lending PSL loans stay classified till repayment/maturity. A de-tag before either is a misclassification — no separate de-tagging clock exists. |
| CL-ID-01 | Para 25 | no finding | Each RE maintains an individual borrower account for its share, so every CLA loan must resolve to a paired (originator, partner) account identity. A loan visible in only one book is the canonical symptom of book divergence — the failure mode recon alone cannot localise. |
| CL-COMMIT-01 | Para 8, 21 (Para 23: transfer only to the identified partner) | no finding | The partner RE's irrevocable back-to-back commitment (Para 21) under the ex-ante arrangement (Para 8) must exist before the loan is disbursed, with the rule-pack (agreement terms) version pinned at commitment. Post-origination partner selection fails the Directions. |
| CL-PSL-03 | Para 7.1 (banks), Para 7.2 (UCBs) | no finding | Aggregate bank-book targets: Total PSL 40% of ANBC/CEOBSE, Agriculture 18%, Micro Enterprises 7.5%, Weaker Sections 12% (UCBs: 60% total). This is an AGGREGATE control over the RE's reported book — the CLA share counts toward it; the engine flags shortfalls on the evidence the RE supplies, it does not estimate a book it cannot see. |
| CL-DISC-01 | Para 35, 36 | no finding | Each RE prominently lists all active CLA partners on its website, and its Notes-to-Accounts disclose aggregate CLA quantum, weighted-average rate, fees, sectors, performance and DLG (quarterly/annual). Evidence-presence control: the engine verifies the artefacts exist and are declared current — content adequacy is the auditor's judgement. |
| CL-DISC-02 | Para 13, 12, 11, 14 | no finding | Document-existence control: the loan agreement discloses role segregation (sourcing/servicing) and the single point of customer interface; interface changes only after prior borrower intimation; grievance provisions present; the CLA agreement segregates responsibilities and defines the info-exchange timeframe. Substantive adequacy of wording is counsel's, not the engine's. |
| CL-ESC-03 | AssureCLA derived control supporting Para 26 (escrow routing) | no finding | All disbursements and repayments route through the escrow account, so the escrow statement series IS the evidence base for financial reconciliation. A :28C: sequence gap or a closing→opening balance break proves movements exist that no statement we hold accounts for. A gap indicates the reconciliation evidence may be incomplete — the status holds until continuity is restored or independently explained. |
| CL-TRF-03 | AssureCLA derived evidence test supporting Para 22 (booking within 15 calendar days) | no finding | 'Booked within 15 days' is four separate events that must each land inside the window: disbursement, partner acceptance, cash reimbursement, and the partner's CORE-ledger GL posting. A middleware acceptance is not a booking. The canonical failure is a GL posting that slid past day 15 through a weekend batch while every upstream message looked on-time. |
Share recomputation per loan
| Loan | Originator share | Retention floor |
|---|---|---|
| cla-50-loan-1 | 20.0000% | no finding |
| cla-50-loan-2 | 20.0000% | no finding |
| cla-50-loan-3 | 20.0000% | no finding |
Synthetic throughout.The lenders are fictional, the loans are generated, and no real arrangement is involved. The engine, the controls and the findings are the product’s own; the data is not. Signals only — the lending decision stays with the lenders.