1. Begin with the service that can fail
A bank can suffer a loss even when borrowers repay and market prices remain stable. A payment may be sent twice, a customer record may be altered without authority, a reconciliation may remain unresolved or a critical service may stop during an outage. Operational risk concerns failures involving people, processes, systems or external events. The conventional Basel definition includes legal risk and excludes strategic and reputational risk, although an operational incident can clearly damage reputation and force strategic changes.
Start with the service and trace how it is delivered. For a customer payment, the chain includes instruction capture, authentication, approval, screening, posting, settlement and confirmation. Each stage depends on staff, data, technology and sometimes external providers. A failure map should identify what can go wrong, who can detect it and how the bank can recover. Merely writing “system risk” against the whole service does not identify a useful control or an accountable owner.
Throughout this chapter, fictional Meridian Bank processes supplier payments for business customers. Its main application is hosted by a technology provider, while branch staff approve exceptional instructions. One incorrect file can therefore involve customer authentication, staff judgement, vendor availability and settlement timing. This case shows why operational risk management is a continuing discipline across departments. Outsourcing a technical component changes the dependency; it does not remove the bank's responsibility for delivering its service and managing the resulting exposure.
2. Classify events without losing their causes
The familiar event classification includes internal fraud, external fraud, employment practices and workplace safety, clients and business practices, damage to physical assets, business disruption and system failures, and execution or process management. These categories organise loss information. They should not replace investigation of the underlying cause. An unauthorised transfer could involve stolen credentials, employee collusion or an incorrect approval process, and the most effective response depends on which mechanism actually occurred.
Internal fraud involves intentional misconduct with an internal party's involvement; external fraud involves an outside party's dishonest activity. Employment-related events concern matters such as workplace obligations and safety. Client and business-practice events can involve unsuitable conduct or failure to meet professional obligations. Physical damage covers relevant destructive events. Disruption and system failure concern interrupted services, while execution and process failures include errors in processing, delivery and record management. Exact reporting classifications should follow the bank's approved taxonomy and applicable requirements.
Do not count the same financial loss several times because it has several causes. Assign a consistent event record, then attach contributing factors and related risk categories. For example, a fraudulent instruction accepted during a system outage may involve both external deception and a weakened manual process. The record should preserve both facts without doubling the loss. Consistent classification supports comparison across branches and periods, but a useful database also retains the narrative needed to improve controls.
3. Separate frequency from severity
Frequency describes how often events occur; severity describes the loss when they occur. A bank may experience many small processing errors and very few catastrophic disruptions. Average annual loss alone can hide this difference. If one process produces ten events averaging INR20,000, its simple annual expected loss is INR2 lakh. Another process may have a small probability of a much larger loss. The response should consider both the routine cost and the potential for disruption beyond the average.
For a simple teaching model, expected annual loss equals expected annual frequency multiplied by mean severity. With twelve events per year and INR50,000 average severity, the result is INR6 lakh. Reducing frequency to eight lowers it to INR4 lakh if severity is unchanged. Reducing severity through faster detection can also help. This multiplication assumes compatible definitions and a suitable relationship between event counts and losses; it is not a complete model of the tail or a regulatory capital calculation.
Severe events can be absent from a short internal history without being impossible. A bank that has never experienced a major data-centre failure still needs resilience arrangements. External event information, scenarios and expert judgement can complement internal data, provided differences in scale and controls are considered. Avoid assigning an event zero probability merely because the observation window contains zero occurrences. Equally, do not copy another institution's loss amount without analysing whether the same mechanism and exposure exist here.
4. Govern risk where the work happens
The board and senior management establish expectations, appetite and oversight. Business and operations teams own the risks in their activities. Independent risk and compliance functions challenge assessments and monitor adherence, while internal audit provides assurance. Operational risk is too widely distributed to be owned solely by a central specialist team. The person responsible for a process must understand its controls, exceptions and recovery arrangements, including dependencies on another department or an external service provider.
A risk and control self-assessment, often called RCSA, helps process owners identify inherent exposure, assess controls and judge residual risk. It should use evidence such as incidents, near misses, audit findings and control tests. A workshop that declares every control effective because a policy exists adds little value. Ask whether the control is suitably designed, consistently performed and capable of detecting the relevant failure in time. Record missing evidence rather than silently treating it as proof of effectiveness.
Responsibilities need to remain clear during change. A new payment channel may alter authentication, volumes, data flows and customer expectations. Product and change approval should assess those effects before launch, establish limits and define heightened monitoring where appropriate. A project team should not disappear at launch without transferring ownership and unresolved issues. The first months of operation can reveal assumptions that testing missed, so post-implementation review is part of sound management rather than an admission that the project failed.
Follow the process
No autoplay. Read each step, or use Play, Pause and Next. Reduced-motion settings remove transitions.
Establish what actually settled.
Review system status and retry controls.
Evidence matters more than a circular.
All steps are available above.
5. Design controls around actual failure modes
Preventive controls aim to stop an event, detective controls identify it, and corrective controls support recovery or remediation. Dual approval can prevent an unauthorised payment; reconciliation can detect a mismatch; a controlled reversal process can correct an eligible error. These layers should fit together. A detective control performed weeks after an irreversible transfer may reveal the problem but do little to limit the loss. Timing is therefore part of control effectiveness, not simply an administrative detail.
Segregation of duties reduces opportunities for error and concealment. The same individual should not be able to originate, approve and reconcile a sensitive transaction without effective independent checks. Access should match responsibilities and be reviewed when roles change. Shared credentials weaken accountability because an audit trail can no longer identify who acted. Emergency access may be necessary, but it needs authorisation, recording, expiry and subsequent review. A permanent emergency exception eventually becomes an uncontrolled ordinary process.
Automation can improve consistency and scale, yet a wrongly configured rule can repeat an error across thousands of transactions. Control design should include input validation, change approval, testing, monitoring and exception handling. Manual fallback arrangements also require controls: an outage should not automatically justify removing every approval step. The practical question is how essential service can continue within acceptable risk when normal systems are unavailable, and which transactions must wait until reliable processing is restored.
6. Use incidents, near misses and indicators
An incident database should capture dates, amounts, recoveries, event types, causes, affected processes and actions. Record gross loss separately from insurance or other recovery so that the underlying exposure remains visible. Discovery date can differ from occurrence date, and accounting recognition can differ from both. Consistent definitions help analysis across periods. Near misses also matter: a duplicated file caught before settlement can reveal the same process weakness as a duplicate payment that actually leaves the bank.
Key risk indicators provide signals about changing exposure or control performance. Examples include overdue reconciliations, failed processing jobs, privileged-access exceptions, staff turnover in critical roles and unresolved high-severity incidents. A threshold needs an owner and response. An indicator that repeatedly turns red without action becomes a decorative dashboard. Explain whether a movement reflects a genuine deterioration, a change in transaction volume or improved reporting, rather than interpreting every increase as equivalent evidence of worsening risk.
Consider unresolved payment breaks increasing from twenty to eighty while transaction volume rises only 10%. The bank should investigate ageing, value, root causes and concentration, not merely celebrate higher throughput. Some breaks may be low-value administrative differences; others may conceal unsettled obligations. A count alone is insufficient. Combining count, monetary value and ageing gives a more decision-relevant picture, and the response should prioritise the cases with the greatest potential customer or financial impact.
Work the cases yourself
7. Build scenarios that expose weak assumptions
Scenario analysis explores plausible severe events that limited historical data may not capture. Start with a coherent narrative, such as a vendor outage during a peak payment day, then identify affected services, duration, financial consequences and customer impact. Challenge the assumed availability of staff, backup systems and communications. A scenario is not useful if every recovery resource is assumed to work perfectly while the primary system fails. The adverse conditions should test the dependencies that matter to the service.
Suppose a four-hour outage affects 8,000 payments. An illustrative direct handling cost of INR150 per delayed payment gives INR12 lakh. Add a separately estimated INR8 lakh of remediation and INR5 lakh of other direct costs to obtain INR25 lakh in the stated scenario. Avoid adding an unsupported reputation multiplier merely to make the result larger. Record unquantified consequences separately, with their potential mechanism. Transparency about uncertainty is more informative than an invented precise estimate of every possible effect.
Scenarios should lead to action and testing. If the case assumes a backup provider, verify capacity, access, data compatibility and the time needed to switch. If manual processing is proposed, estimate how many transactions staff can safely handle. Recovery actions should be feasible within the outage, not just documented in a plan. Reverse scenarios can ask what combination would exceed the bank's service tolerance or available resources, helping management identify the most important improvements before an incident occurs.
8. Resilience means sustaining critical operations
Business continuity and disaster recovery are related but distinct. Continuity concerns maintaining or restoring important activities; technology recovery is one component. Staff availability, premises, communications, third parties and data integrity can all constrain the service. Recovery time objectives describe intended restoration time, while recovery point objectives concern tolerable data loss in time terms. These objectives must be interpreted within the bank's approved framework and tested against the needs of the service, rather than selected merely because a vendor advertises them.
A backup is not sufficient evidence of recoverability. The bank should test restoration, access, integrity and reconciliation after recovery. An application may restart while transaction records remain incomplete or duplicated. Returning to service without checking the data can transform an outage into a financial incident. Exercises should include decision making and communication, not only a technical switch. The people authorised to invoke the plan need reliable information and substitutes if they themselves are unavailable during the disruption.
Operational resilience focuses attention on critical operations and their dependencies under disruption. A service can depend on several systems that share one power source, network path or provider, creating hidden concentration. Mapping dependencies reveals these common points of failure. Tests should progressively challenge the mapping and recovery assumptions. A successful exercise is valuable when it produces evidence of capability and closes weaknesses; a scripted demonstration that avoids difficult conditions provides much less assurance about actual resilience.
9. Insurance and capital do not replace controls
Insurance may transfer eligible financial consequences, but coverage has limits, exclusions, deductibles and claim conditions. A policy can leave interruption costs or customer consequences uninsured, and payment may arrive long after the event. The bank must preserve evidence and meet notification obligations. Insurance should be analysed as a conditional recovery source, not an immediate reduction in every loss estimate. A claim dispute can itself create legal and operational work that needs time and resources.
For a simple illustration, a covered gross loss of INR50 lakh with an INR5 lakh deductible and INR30 lakh policy cap produces a maximum model recovery of INR30 lakh and retained loss of INR20 lakh. This assumes the policy pays the eligible amount above the deductible up to the cap; actual wording may differ. A recovery estimate should not exceed the eligible loss. The calculator makes the order explicit so that students can see why changing a deductible or limit changes retained exposure.
Operational-risk capital provides loss-absorbing capacity under the applicable framework; it does not prevent an incorrect transfer or restore an unavailable system. Historical approaches, current domestic rules and newer global approaches must be distinguished. Chapter B03 discusses the retrieved domestic capital method and its version limitation. Here the focus is management: strengthen controls, monitor exposure, prepare recovery and retain adequate resources. A bank should not justify an avoidable process weakness simply by pointing to capital held against operational risk.
10. Integrate risks through a common scenario
Integrated risk management connects risks across business lines and categories so that management can assess the institution as a whole. An outage can delay collections, trigger customer withdrawals, require emergency funding and force asset sales. Those effects cross operational, liquidity and market risk. If departments assess only their own category, they may miss the combined demand on cash and capital. Integration requires consistent data, definitions, governance and a narrative that explains how the effects arise together.
Aggregation is not always simple addition. Some losses overlap, some offsets are unavailable and some effects amplify one another. A compensation payment recorded as operational loss should not be counted again as a separate reputation loss without identifying an additional consequence. Conversely, a funding cost caused by withdrawals can be a distinct effect. Document the links and prevent double counting. Correlation-based models can support aggregation, but their assumptions need challenge where stress creates nonlinear feedback or shared dependencies.
Integration also faces organisational challenges. Systems may use different counterparty identifiers, dates and currencies, while departments may resist sharing inconvenient information. A common data dictionary and accountable reconciliation process are therefore practical risk controls. Senior management should receive a consolidated view that preserves important local weaknesses. A comfortable group total can conceal a shortfall in a subsidiary or currency where resources cannot be transferred quickly. The objective is a decision-ready picture, not a single reassuring number.
11. Review a payment incident from end to end
Meridian receives a duplicate supplier file after a network interruption. An employee assumes the first submission failed and approves the second without checking settlement status. The event combines an ambiguous system response, a weak retry procedure and inadequate independent verification. The immediate priority is to establish what actually settled, prevent further duplication and use authorised recovery channels. Blaming the employee alone would overlook the process and system conditions that made the mistake likely to recur.
The review should ask why the system did not provide a unique transaction reference, why the retry rule was unclear and why the checker could not see the first payment's status. Corrective actions may include idempotent processing, clearer status messages, independent reconciliation and targeted training. Each action needs an owner and a test. Training alone is unlikely to fix a system that continues to display ambiguous information at the critical decision point.
For revision, connect event classification, frequency and severity, controls, indicators, scenarios, resilience and integration. Explain both the immediate response and the preventive improvement. A strong answer distinguishes gross loss from recovery, a policy from an operating control and a backup from tested recovery. The practical aim is reliable service under ordinary and adverse conditions, supported by evidence and clear responsibility. That is why operational risk management belongs in daily banking decisions rather than only in an annual compliance exercise.
25 original questions
Choose one best answer. Explanations address the answer and distractors; no negative marking.
Browser-local saving is optional.
Ready: 25 questions.
Sources and version notes
- IIBF CAIIB Rules and Syllabus 2026: Module B, PDF pages 8 to 9. B01 to B08 are editorial IDs matching the eight topic groups.
- BCBS revised operational-risk principles: 31 March 2021 revision; older 2011 version is superseded.
- RBI commercial-bank capital directions: Readable consolidation displayed 1 July 2026. Official index indicated 7 October 2026 update, not retrieved in full. Numeric rule card and BIA discussion are dated to the retrieved text; current consolidation review remains required before academic approval.
Prepared 7 November 2026. Review edition; independent academic approval and current capital-consolidation verification remain pending. Original examples are illustrative. External source links require internet.