Training must be supported by checks of actual conduct
Paragraphs 43–44 of RBI’s Commercial Banks card directions require random checks and mystery shopping to test agent briefing and awareness, prevent credit-card mis-selling through incomplete or incorrect pre-issuance information, and record repetitive complaints with appropriate action including blacklisting. The issuer remains liable for its agents’ acts. This guide uses the directions updated to 1 October 2026, checked on 27 October 2026. It applies to the specified commercial-bank population, excluding Small Finance Banks, Payment Banks and Local Area Banks. RBI Commercial Banks card directions: paragraphs 3 and 43–44.
Compare the pitch with the applicable product terms
Retain the card and offer version, approved disclosures and the statement actually made in the tested interaction. Inspect fee, eligibility, interest, reward and other material statements relevant to the case rather than scoring the friendliness of the conversation. This is a proposed test design, not an RBI-prescribed sample size or score.
Use both random and risk-informed evidence
Random checks and mystery shopping can reveal conduct that does not appear in a training attendance sheet. Link test findings to the agent, agency, product and channel. A targeted sample can investigate a known concern, while its result should not be presented as an unbiased estimate of all sales. Preserve the population and selection basis.
Do not lose repeated complaints through case closure
Maintain an agent-linked complaint history while distinguishing repeat contacts on one complaint from separate complaint events. Investigate the alleged statement and customer outcome. Closing an individual ticket should not delete a recurring pattern from the monitoring file. The rule calls for appropriate action; it does not supply a universal number of complaints that must be reached before any action is possible.
Retest the underlying conduct
Document the investigation, corrective training, agency action and other appropriate measures, then test whether the problem recurs. A dedicated helpline and email-id must be available for complaints of mis-selling or harassment by issuer representatives. This article does not invent a fixed compensation tariff, automatic debt waiver or an RBI-approved acceptable mis-selling percentage.
Applicable primary references
RBI Commercial Banks card directions: paragraphs 3 and 43–44
Six worked illustrations
All facts and amounts below are fictional. Assumptions describe the model, not additional statutory rules.
Illustrative example 1: Mystery-shop sample
Forty tested sales interactions include six material departures from the verified terms. The observed exception rate is 15%. It describes this sample, not a regulatory tolerance or a statistically proven population rate.
Illustrative example 2: Training is not the test outcome
Of 50 agents, 45 have training attendance evidence, or 90%. A separate ten-agent conduct test finds three exceptions. Do not report 90% attendance as proof that 90% of sales comply; the tested conduct exception rate is 30% for that separate sample.
Illustrative example 3: Repeated contacts versus complaints
Twelve contacts concern seven distinct complaints against an agent, with five contacts repeating existing cases. Report seven complaint events and twelve contacts. Preserve the event history when assessing repetitive complaints rather than counting every callback as a new complaint.
Illustrative example 4: Overlapping failures
Among 30 interactions, eight misstate a fee and five misstate a benefit; three do both. Ten unique interactions have an exception: 8 + 5 − 3. The 13 failure occurrences should not be presented as 13 affected interactions.
Illustrative example 5: Fee impact working
Assume five customers were incorrectly told a verified ₹500 fee would not apply, and each was actually charged once. The disputed charge population is ₹2,500. This identifies amounts to investigate; it does not determine an automatic compensation entitlement.
Illustrative example 6: Retest coverage
Eight corrective actions are assigned. Six have completed evidence and two are pending. Completion is 75%. If four of the six completed items are retested, retest coverage of completed items is 66.67%, rounded to two decimals, rather than 75%.
Ten-step verification table
| Step | Action | Evidence |
|---|---|---|
| 1 | Confirm issuer and directions | Applicable population and version |
| 2 | Retain approved product terms | Product and offer version |
| 3 | Map sales agents and channels | Agency and employee identifiers |
| 4 | Design random checks and mystery shops | Population, selection and test criteria |
| 5 | Capture actual representations | Interaction evidence |
| 6 | Compare against verified terms | Specific incomplete or incorrect statements |
| 7 | Maintain complaint history | Unique events and repeat contacts |
| 8 | Investigate agent-linked patterns | Evidence and customer impact |
| 9 | Take and document appropriate action | Training, agency action and resolution |
| 10 | Retest and preserve outcomes | Verified correction and recurring issues |
Sixteen practitioner FAQs
1. Is a training attendance sheet sufficient monitoring?
No. Paragraph 43 also requires a system of random checks and mystery shopping.
2. Does this rule prescribe the sample sizes in the article?
No. They are fictional control examples.
3. What is the focus of the mis-selling test?
Incomplete or incorrect pre-issuance information compared with the verified product terms.
4. Can a friendly sales interaction still fail?
Yes. Accuracy and completeness must be assessed separately from manner.
5. Is the issuer responsible for agent conduct?
Paragraph 44 states issuer liability for agents’ acts.
6. Must repetitive complaints be recorded?
Yes, with appropriate action as required by the provision.
7. Is a fixed three-complaint trigger prescribed here?
No such numerical threshold is supplied by the cited paragraphs.
8. Should repeated phone calls be counted as separate complaints?
Distinguish contacts from unique events while retaining both histories.
9. Does closing a ticket erase a pattern?
The agent-linked monitoring history should preserve relevant recurring issues.
10. Can a targeted sample represent every sale statistically?
Do not make that inference without a sound sampling basis.
11. How should overlapping findings be reported?
Show both unique affected interactions and defect occurrences.
12. What channels must exist for these complaints?
A dedicated helpline and email-id for mis-selling or harassment complaints as specified.
13. Is blacklisting the only possible action?
The provision refers to appropriate action including blacklisting; assess the actual facts and required response.
14. Does a disputed fee automatically create a fixed award?
No. Determine actual redress under the applicable facts and law.
15. What should a retest examine?
Whether the specific inaccurate or incomplete representation has been corrected in practice.
16. What is the final evidence pack?
Approved terms, test records, complaint history, investigation, action and retested outcomes.
Conclusion
Test whether sales agents explain card terms accurately, retain repeat-complaint history and connect monitoring results to documented corrective action.