Collection Call Audit for NBFCs: Scorecard and RBI Rules
Last reviewed: November 2026
A collection call audit for NBFCs is a structured review of recorded recovery calls to check that employees and agency callers followed RBI rules on calling hours, language, threats, disclosures and dues accuracy. It is how a lender proves that its recovery effort stays firm without turning into harassment.
Heads of collections and chief compliance officers at non-banking financial companies (NBFCs) will find a findings matrix, a weighted scorecard, the Reserve Bank of India (RBI) rules in force now and from 1 January 2027, and a sampling formula here. We also explain how Digiqt's artificial intelligence (AI) development team can screen every call.
At a Glance
- Rules tighten on 1 January 2027. Call recording, a single 08:00–19:00 window and a defined list of harsh practices arrive with the Third Amendment.
- Use fatal errors. Abuse, threats and false legal claims should fail a call outright.
- Sample with a formula, then add every complaint call and every machine-flagged call.
- Agency calls are your calls. Outsourcing does not move the responsibility.
Remove, Retrain or Warn: How Should NBFCs Act on Collection Call Audit Findings?
NBFCs should act on collection call audit findings by mapping each one to a fixed action, owner and deadline, so the same breach gets the same response in every branch and agency. This illustrative matrix is a starting point.
| Finding | Action | Owner | Deadline |
|---|---|---|---|
| Abusive language or threat | Remove the recovery agent from collections; inform agency; check borrower complaint | Head of collections | 24 hours |
| Dues discussed with a relative | Retrain; warn; review borrower contact data | Compliance | 3 days |
| Out-of-hours call | Fix dialler time rules; warn caller | Collections operations | 3 days |
| Wrong dues stated | Correct data feed; call borrower back | Collections operations | 2 days |
| Missing recording disclosure | Update script; refresher training | Quality team | 7 days |
| Agency fails audit two months running | Review contract and recovery agent incentives | Vendor management | 30 days |
Out-of-hours breaches usually trace back to collections dialler list management, so fix calling windows at the source. Also confirm each agency caller's debt recovery certificate through debt recovery agent (DRA) certification tracking.
What Is a Collection Call Audit for NBFCs?
A collection call audit for NBFCs listens to, scores and acts on a sample of recovery calls made by in-house teams and collection agencies. It checks conduct, not just results, which is why it sits beside collection agency performance management rather than inside it.
Is a Collection Call Audit Different From a Sales Call Audit?
A collection call audit is different from a sales call audit because it tests conduct under RBI recovery rules, not product and pricing disclosures. Sales calls follow the checks in our guide to telesales script compliance for lenders.
Keep the two scorecards, samples and owners separate. A firm recovery caller and a persuasive sales caller can both pass, while failing for very different reasons.
Which Scorecard Points Decide Whether a Recovery Call Fails?
A recovery call fails on a collection call audit scorecard of eleven weighted parameters adding to 100, with four of them acting as fatal errors. Paste this illustrative scorecard into your quality form.
| No. | Parameter | Weight | Fatal? |
|---|---|---|---|
| 1 | Call within permitted hours (or borrower's requested time) | 10 | Yes |
| 2 | Caller states own name, the lender's name and the purpose | 10 | No |
| 3 | Right-party check before discussing dues | 10 | Yes |
| 4 | Civil, non-abusive language throughout | 15 | Yes |
| 5 | No threats, intimidation or false legal claims | 15 | Yes |
| 6 | Dues, charges and due dates stated accurately | 10 | No |
| 7 | Recording disclosure given (mandatory from 2027) | 5 | No |
| 8 | Only official payment channels offered | 10 | No |
| 9 | Grievance contact shared when asked | 5 | No |
| 10 | Hardship or sensitive occasion handled with care | 5 | No |
| 11 | Promise to pay and next step logged correctly | 5 | No |
Set a pass mark, such as 80, and fail any call with a fatal error regardless of its score.
Is your audit team hearing only a few per cent of recovery calls?
Explore an AI Collection Call Audit Build With Digiqt
Share a sample of recordings and your scorecard, or plan a call-audit model with our AI team.
What Do RBI Rules Say About Recovery Calls?
RBI rules on recovery calls change on 1 January 2027, when the NBFC Responsible Business Conduct Third Amendment Directions, 2026 (dated 6 August 2026) delete paragraphs 89 to 100 and add Section J.
| Topic | Until 31 December 2026 | From 1 January 2027 |
|---|---|---|
| Calling hours | Para 100: no calls "before 8:00 a.m. and after 7:00 p.m." for overdue loans; microfinance para 91(2): not before 9:00 a.m. or after 6:00 p.m. | Para 100Y(4): contact "only between 08:00 hours and 19:00 hours" unless the borrower asks otherwise |
| Harassment | Para 100: no "intimidation or harassment of any kind" | Para 100Z: eight deemed harsh practices, including "minatory or abusive language," threatening or anonymous calls and "false or misleading representations" |
| Who to discuss dues with | Para 91(3) bars harassing relatives, friends or co-workers (microfinance) | Para 100Y(1): discuss dues only with the borrower or guarantor |
| Recording | No specific rule | Para 100P: record call content, log times and numbers, keep for six months, tell the borrower |
| Incentives | No specific rule | Para 100Q: targets must not induce harsh practices |
| Grievance details | General grievance rules | Para 100AA: every recovery communication carries the grievance officer's name, email, phone and address |
| Telecom rules | – | Para 100AB: comply with the Telecom Regulatory Authority of India (TRAI) Telecom Commercial Communications Customer Preference Regulations (TCCCPR), 2018 |
Paragraph 100Y(6) also asks callers to avoid inappropriate occasions such as bereavement, medical emergencies, calamities or marriages.
Which Recovery Call Timings Breach the Rules?
Recovery call timings breach the rules when a call to a non-microfinance borrower falls before 08:00 or after 19:00, or a call to a microfinance borrower falls outside 09:00–18:00 under the current paragraph 91. Test these sample call logs against your dialler data.
| Call Log Entry | Until 31 December 2026 | From 1 January 2027 |
|---|---|---|
| Vehicle loan, 7:40 a.m., no borrower request | Breach | Breach |
| Personal loan, 7:20 p.m., no borrower request | Breach | Breach |
| Microfinance loan, 6:30 p.m. | Breach under para 91(2) | Test against the 08:00–19:00 window in para 100Y(4) |
| Any loan, 9:00 p.m., borrower asked for a late call and it is logged | Check policy | Permitted only with the borrower's express request or authorisation |
| Twelve calls in one day to the same borrower | Persistent calling, a breach | Excessive calling under para 100Z(4) |
Out-of-hours rate = calls outside the permitted window ÷ total calls × 100. In an illustrative month, 240 of 60,000 calls fall outside the window, an out-of-hours rate of 0.4% that should be traced to the recovery agent or dialler campaign behind each one.
How Many Recovery Calls Can a Manual Audit Team Cover?
A manual audit team covers only a few per cent of recovery calls, which leaves most breaches unseen. In this illustrative setup, an NBFC makes 60,000 collection calls a month and employs four call auditors.
| Measure | Manual Listening | AI Screening Plus Human Review |
|---|---|---|
| Calls reviewed per auditor per day | 40 | 40 deep reviews of flagged calls |
| Calls reviewed a month (4 auditors, 22 days) | 3,520 | 60,000 screened; 3,520 reviewed |
| Coverage | 5.9% | 100% screened |
| Out-of-hours calls found | Only those in the sample | All, from call logs |
At a 1% breach rate, 600 bad calls happen a month, and random manual sampling catches roughly 35 of them.
How Can AI Help Audit Every Collection Call?
AI can help audit every collection call by converting speech to text, checking call logs against permitted hours and flagging risky phrases for human review. Digiqt's custom AI development team runs a discovery sprint, assesses your data, builds and validates a custom model, then deploys it with continuous improvement.
Each audit step changes like this once a custom model built by Digiqt is in place:
- Coverage: random sample only → every call screened.
- Calling hours: checked from sampled calls → every call log tested against the window.
- Abuse and threats: found if the auditor hears it → phrases flagged with the reason shown (explainable AI).
- Disclosures: ticked by the auditor → missing name, purpose or recording disclosure flagged.
- Reporting: monthly spreadsheet → real-time and batch analytics by agency and caller.
Explainable flags matter because a caller facing disciplinary action, and the compliance team defending it, both need to see the exact words that triggered the finding.
How Many Calls Should an NBFC Audit in a 60,000-Call Month?
An NBFC with 60,000 calls in a month should audit about 382 calls at random, stratified across agencies and delinquency buckets, plus risk-based additions. The formula below sets that number.
Sample size n = z² × p × (1 − p) ÷ e², then adjust for population: n(adj) = n ÷ (1 + (n − 1) ÷ N)
Here z is 1.96 for 95% confidence, p is the expected breach rate (0.5 if unknown), e is the margin of error and N is the month's calls.
In this illustrative month, N is 60,000 calls, so n = 1.96² × 0.5 × 0.5 ÷ 0.05² = 384.2, and n(adj) = 384.2 ÷ (1 + 383.2 ÷ 60,000) ≈ 382 calls. Split them in proportion to call volume across agencies and buckets from your bucket-wise collection strategy.
Then add three risk-based layers:
- Every call linked to a borrower complaint.
- Every call logged outside 08:00–19:00.
- Every call from a caller who failed last month's audit.
Feed borrower complaints from your loan grievance redressal system into the sample. Treat recordings as personal data under Digital Personal Data Protection (DPDP) Act compliance for lenders, use collections workforce planning to size the audit team, and register message templates through distributed ledger technology (DLT) template registration.
Want explainable flags on every abusive or out-of-hours recovery call?
Start a Collection Call Audit AI Discovery Sprint
Begin with one agency's recordings, or visit Digiqt to request a consultation on recovery call quality.



