PAR Spike, Ghost Collections and a Star Officer
At the end of Q2, the Kalmeshwar branch of an NBFC-MFI reports 3,900 active borrowers, PAR-30 of 6.4% (up from 2.1% in Q1), PAR-60 of 2.9%, and collection efficiency of 97.8% - the best in the region. Internal audit finds that credit officer R, who manages 512 borrowers across 41 joint-liability groups (branch average: 380 borrowers), has (i) collected ₹2.4 lakh from 11 people who are not borrowers and shown it against 11 delinquent accounts; (ii) reclassified 9 defaulted loans as fresh loans in the MIS to reset ageing; and (iii) himself signed as witness on 6 loan documents. In the same branch, 3 groups include 18 members declaring household income of ₹3.4 lakh, and 5 of those members hold loans from five lenders. Unseasonal hail in August affected 42% of R's borrowers. The regional head wants R 'handled quietly' because he is the best collector in the region.
1. Read the numbers: what do PAR-30, PAR-60 and 97.8% collection efficiency together tell you, and why is the collection efficiency figure unreliable?
Read the worked answer
The buckets are inconsistent with the collection efficiency. On 3,900 borrowers, PAR-30 of 6.4% is about 250 accounts and PAR-60 of 2.9% is about 113. Roughly 45% of everything that enters the 30+ bucket is rolling into 60+, which means the early bucket is not being cured - the branch is not recovering marginal cases, it is losing them. A healthy branch cures most of its 30-day bucket within a month; a flow rate of 45% into 60+ is the signature of either a real income shock or of numbers that are not what they seem.
The 97.8% is an artefact of the fraud. Collection efficiency is collections divided by collectibles. Remove ₹2.4 lakh of ghost collections from non-borrowers and restore the 9 evergreen reclassifications to their true days past due, and both the numerator and the denominator change: cash is credited to accounts that are still bad, and 9 accounts have been removed from the delinquent base altogether. Real collection efficiency is materially lower.
The diagnostic point. A branch that reports the region's best collection efficiency while its PAR doubles is not a high-performing branch; it is a branch whose MIS has been manipulated. The divergence itself is the control that should have caught this at head office - PAR and collection efficiency moving in opposite directions is a classic fraud signal and should be an automatic exception report. The hail that hit 42% of R's borrowers is also relevant: it should have produced a documented restructuring decision for the affected groups, not concealment of the resulting arrears.
2. Set out your actions on the officer, the accounts and the branch in the next 30 days, including classification and provisioning implications.
Read the worked answer
Officer. Suspend R immediately, preserve his device, collection app logs and any manual receipts, and run a forensic review of all 512 accounts and of other officers with similar patterns. Separate the conduct issue from the performance issue: a good collector who fabricates entries is a fraud risk, not an asset, and 'handling it quietly' converts a control failure into a governance failure. If the ₹2.4 lakh is not recoverable, file an FIR and place the matter before the audit committee.
Accounts. Reverse the 9 reclassifications to their true days past due and restate the branch PAR. Return the ₹2.4 lakh to the 11 non-borrowers against receipts with a written apology - this is money taken from third parties, and returning it limits legal exposure and community damage. Post the 11 corresponding loans to their correct buckets, which will create fresh NPAs and require provisioning at the higher rates applicable on classification and, if the board's policy so provides, a fraud provision.
The 18 high-income members. Households earning ₹3.4 lakh do not meet the ₹3 lakh income test, so those loans cannot be counted as microfinance loans. Reclassify them and test at company level whether the requirement that microfinance loans form at least half of the total portfolio is still met - if not, the misdeclaration has a regulatory consequence far larger than one branch.
The 5 multi-lender borrowers. Pull bureau reports, stop further lending, and assess whether the combined obligation breaches the 50% household repayment cap.
Hail-affected accounts. Apply the board's restructuring policy, extend tenure where justified, and document it - that is the legitimate route that R bypassed.
3. Identify the credit appraisal failures the audit has exposed and the three controls you would change.
Read the worked answer
What failed
- Household income was never properly assessed; declarations of ₹3.4 lakh were accepted for a product with a ₹3 lakh ceiling.
- No effective credit bureau check, given five lenders on a single borrower.
- Documentation was witnessed by the same officer who originated and collected the loan - no independent verification anywhere in the chain.
- Span of control (512 clients, 41 groups) exceeded the branch average by a third, which is precisely the condition under which ghost entries become possible.
- The MIS permitted back-dated reclassification with no maker-checker.
- Group quality was weak, with overlapping membership that lets one person's default contaminate several groups.
Three controls to change
- Immutable ageing. Days past due computed by the system from the disbursement date, with no manual override; any reclassification must be made by operations under maker-checker with a mandatory reason code and an audit trail. Fraud in microfinance almost always starts with a date.
- Collections only through traceable channels. Borrower-initiated UPI or account transfer, or cash with a numbered receipt reconciled daily, plus independent call-backs to a 5% random sample of borrowers each week confirming amounts received. Ghost collection cannot survive a call-back.
- Hard stops in the loan management system. No sanction is generated without a documented income assessment below ₹3 lakh, a bureau pull for all adult household members, a combined repayment obligation within 50% of household income, a lender-count flag, and an independent witness drawn from outside the officer's own portfolio. Add a span-of-control cap of 400 clients and a quarterly audit of 10% of groups.