Last updated · By the Clerk2Officer editorial team · Report a correction
Bank computerisation map
Capture instructions through controlled channels.
P2C01: Essentials of Bank Computerisation
1. What computerisation changes in a bank
Bank computerisation uses information systems to record, process, store, retrieve and communicate banking information. The important change is not the replacement of a paper ledger by a screen; it is the ability to apply consistent rules, connect related records and make information available across authorised service points. A deposit transaction affects the customer balance, cash or settlement position, accounting records and audit trail. If these effects are disconnected, a computer may merely reproduce the weaknesses of a fragmented manual process more quickly.
- Computerisation connects records, processing rules and authorised access to information.
- A banking transaction has customer, accounting and control consequences.
Early automation often addressed individual tasks such as posting a ledger or preparing a statement. Branch computerisation brought more functions together within a branch, while networked core banking enabled a bank-wide service environment. These are conceptual stages rather than a claim that every Indian bank adopted identical equipment in the same year. A student should understand the progression from isolated processing to integrated information, because it explains both improved customer access and the increased consequences of a common-system failure.
- Isolated task automation, branch integration and bank-wide core banking are different stages.
- Shared infrastructure increases both service reach and the importance of resilience.
The main benefits include faster processing, consistent calculation, easier retrieval, broader service channels and more timely management information. However, automation faithfully applies incorrect parameters too. If the wrong interest basis is entered for a product, thousands of accounts may be affected rather than one ledger. Computerisation therefore changes the nature and scale of control: input quality, authorisation, configuration, reconciliation and exception handling remain banking responsibilities even when most calculations occur automatically.
- Automated processing can spread a configuration error across many accounts.
- Input validation and control of parameters remain essential banking functions.
2. Components: hardware, software, data and people
Hardware includes physical equipment such as workstations, servers, storage devices, scanners and network components. Software includes operating systems, database-management systems, applications and security tools. Data contains the customer, account, transaction and reference records those applications use. People and procedures determine who can enter, approve, change or review them. A working bank system requires all these elements; installing powerful servers does not compensate for missing approval rules, poorly trained users or unreliable source records.
- Hardware is physical equipment; software supplies instructions and application functions.
- Data, people and procedures are necessary parts of the banking information system.
The processor executes instructions, working memory supports active processing and persistent storage retains information beyond the current session. RAM should not be confused with the long-term record of customer accounts. A workstation may lose unsaved local work when power fails, while a properly committed transaction remains in the bank’s persistent database. Whether a transaction committed must be established from system records; the disappearance of a screen is not evidence that the bank did or did not post the transaction.
- Working memory and persistent storage perform different roles.
- Transaction status must be verified from records rather than inferred from a screen failure.
Application software implements business functions such as account opening, deposits, loans and payments. The database-management system manages structured records, relationships, access and transaction processing. An operating system manages the underlying machine resources. These layers are related but not interchangeable. For example, a slow account enquiry may arise from network delay, application processing or database access; replacing a monitor cannot resolve each possible cause. Correct diagnosis depends on identifying where the failure or bottleneck occurs.
- Operating system, database and banking application are separate functional layers.
- Troubleshooting should identify the affected layer before selecting a remedy.
3. Data organisation and transaction integrity
A field holds one item, such as an account number; a record brings related fields together; a table stores records of a defined kind. Relationships connect customers with accounts and accounts with transactions. A unique identifier helps distinguish customers who share a name and prevents accidental mixing of records. It does not itself prove that two identifiers represent two different people: duplicate customer records can still arise when onboarding controls fail. Data quality includes accuracy, completeness, consistency and timely updating.
- Fields, records and tables organise data at different levels.
- Unique identifiers aid linkage but do not eliminate duplicate-customer risk.
Consider an internal transfer of ₹5,000 from one customer account to another. The system must avoid recording the debit without the corresponding credit as a completed transfer. Transaction controls aim to keep the related changes together, preserve valid accounting relationships, manage concurrent activity and retain committed results. The familiar ACID concepts describe atomicity, consistency, isolation and durability. They help explain reliable database transactions, but they do not replace reconciliation between separate applications or external payment participants.
- A completed internal transfer requires coherent debit and credit effects.
- ACID concerns database-transaction properties; cross-system reconciliation remains necessary.
Atomicity means the defined transaction is treated as a whole; consistency preserves the system’s specified valid states; isolation controls interaction between concurrent transactions; durability preserves a committed result. Suppose two channels attempt withdrawals against the same available balance. Appropriate concurrency controls must prevent both from independently assuming that the full pre-withdrawal balance remains available. The precise implementation differs by platform, but the banking outcome is clear: simultaneous access should not create an unauthorised duplicate use of funds.
- Atomicity, consistency, isolation and durability address different reliability requirements.
- Concurrent requests must be controlled against the actual available funds.
4. Networks within and beyond the branch
A local area network connects devices over a relatively limited location, such as a branch or office. A wide area network connects geographically separated locations, including branches and data centres. The internet is a network of networks used for many public and private services; it is not synonymous with the bank’s internal network. A branch can use controlled connectivity over different carriers and technologies while still applying bank-specific routing, authentication and security arrangements.
- A LAN serves a limited local area; a WAN connects separated locations.
- Internet connectivity and the bank’s internal network are not identical concepts.
A switch commonly connects devices within a local network, while a router directs traffic between networks. Modern equipment may combine functions, so the conceptual distinction should not become a claim that every physical box has only one capability. A wireless access point provides wireless connectivity to a network; it does not itself establish permission to view every banking application. Network attachment, user identity, application authorisation and transaction approval are different stages of access control.
- Switching connects local devices; routing connects networks.
- Being connected to the network does not authorise every application or transaction.
Bandwidth describes the amount of data a link can carry per unit time, while latency concerns delay. A high-bandwidth link can still feel slow when each application action requires many delayed round trips. Throughput is the rate actually achieved, which may be lower than nominal capacity because of congestion, overhead or system limitations. Availability asks whether the service is usable when needed. These measures answer different questions and should not be compressed into the single word “speed”.
- Bandwidth is capacity, latency is delay and throughput is achieved transfer rate.
- Availability is a separate measure of whether the service can be used.
Assume a file contains 20 megabytes, with one byte equal to eight bits, and a link provides an effective 10 megabits per second. Ignoring additional overhead, the file contains 160 megabits and takes 16 seconds to transfer. If actual throughput is only 5 megabits per second, the estimate becomes 32 seconds. This example states decimal units deliberately; confusing megabytes with megabits would understate the transfer requirement by a factor of eight.
- Under the stated units, 20 MB equals 160 Mb.
- Transfer time is 16 seconds at 10 Mbps or 32 seconds at 5 Mbps.
5. Topology, separation and alternate connectivity
Network topology describes how connections are arranged. In a star arrangement, local devices connect through a central point; failure of that point can affect many devices unless redundancy is provided. Mesh arrangements provide multiple interconnections and possible alternate paths, with greater complexity. Physical layout and logical traffic separation need not be identical. Devices attached to the same infrastructure can be separated into controlled segments so that a visitor’s wireless connection does not automatically share the privileges of a teller workstation.
- Topology affects failure points, alternate paths and management complexity.
- Logical separation can restrict traffic even on shared physical infrastructure.
Redundancy is effective only when alternatives do not share the same critical weakness. Two network links following the same damaged cable route may fail together. Two power supplies connected to the same failed source may not keep a service running. A branch continuity design should therefore identify dependencies rather than merely count duplicate components. Alternate connectivity also requires tested routing, sufficient capacity and approved security; an employee’s personal hotspot is not automatically an authorised substitute for the banking network.
- Duplicate components can still share a common point of failure.
- Alternate connectivity must be tested, adequately sized and authorised.
A firewall applies rules to traffic at a network boundary or other control point. Encryption protects data under the relevant cryptographic arrangement, and authentication helps establish who or what is communicating. These controls address different risks. An encrypted connection can still carry an unauthorised instruction from a compromised account, and a firewall cannot decide whether a customer has economically justified a loan. Technical controls support, rather than replace, business authorisation and transaction monitoring.
- Firewall rules, encryption and authentication have different purposes.
- Technical connectivity controls do not replace business-level approval.
6. Uninterruptible power supply and environmental support
An uninterruptible power supply, or UPS, provides temporary power support when the normal supply is interrupted and may also condition the supplied power according to its design. Its purpose is to bridge a disturbance, allow an alternate source to take over or support controlled shutdown. It is not an unlimited source of electricity. Batteries have finite capacity, and runtime depends on load, battery condition, efficiency and the equipment’s actual operating characteristics.
- UPS support is temporary and depends on usable energy and connected load.
- A UPS can support continuity or orderly shutdown; it does not promise unlimited operation.
For a simplified calculation, assume a battery system has 1.2 kilowatt-hours of usable energy after all stated losses and supplies a constant 0.6 kilowatt load. Estimated runtime is 1.2 divided by 0.6, or two hours. If the load rises to 0.8 kilowatt, estimated runtime becomes one and a half hours. These are teaching estimates, not a product specification. Real sizing must account for manufacturer discharge characteristics, reserve capacity, ageing, temperature and the critical equipment to be supported.
- Assumed runtime is usable energy divided by constant load.
- The example gives two hours at 0.6 kW and one and a half hours at 0.8 kW.
A generator and UPS can complement each other: the UPS can support equipment during the time an alternate power source becomes ready, while that source provides longer-duration supply subject to fuel and operating conditions. Environmental controls also include cooling, fire detection, water protection and controlled physical access. A server with electricity but inadequate cooling may still fail. Maintenance and realistic load testing are therefore as important as installing equipment and noting its name in a branch asset register.
- Power continuity includes transition time, fuel, maintenance and tested load support.
- Cooling, fire, water and physical-access risks also affect service availability.
7. Core banking: a shared operating foundation
A core banking solution supports principal banking records and processing through an integrated environment accessible to authorised branches and channels. Its central feature is a consistent bank-wide operating view, not a promise that every function runs on one physical computer. Deposit, loan and customer information may be served through a resilient architecture with multiple components. The customer can receive services beyond the original branch where the bank’s product, access and authorisation arrangements permit them.
- CBS provides an integrated operating foundation for principal banking functions.
- Logical integration does not require one physical computer or unrestricted access.
A customer identifier links the person or entity to relevant relationships, while an account number identifies a particular account. One customer can have several accounts with different holders, mandates, products or legal capacities. A consolidated customer view helps service and risk management, but it must not erase these distinctions. For example, knowing that one person operates both a personal account and a trust account does not make their balances interchangeable or permit unrestricted adjustment between them.
- Customer identity and account identity are distinct.
- A consolidated view must preserve account ownership, mandate and legal capacity.
Core banking usually connects with other systems, such as card processing, payment gateways, treasury, anti-money-laundering monitoring and customer communication. An interface passes defined information between components. An application programming interface is one way of exposing controlled functions or data to another system; it is not a general licence to access the database. Interface controls must address authentication, message validation, duplicate requests, error handling and reconciliation so that each component agrees on the outcome.
- CBS integrates with specialised systems through controlled interfaces.
- Interface success requires validated, authorised and reconcilable messages.
Suppose a mobile transfer request times out after the core system has committed the debit and credit. The customer may see no confirmation even though the accounting event occurred. Repeating the instruction without checking its reference can create a duplicate unless the design recognises and safely handles retries. Staff should use the approved transaction enquiry and reconciliation process, not assume that a timeout means failure. C02 develops the operational stages of posting, authorisation and exception handling in more detail.
- A communication timeout does not establish that the transaction failed.
- Safe retries and enquiry rely on transaction references and recorded status.
8. Availability, capacity and migration
Availability is commonly expressed as usable service time divided by the relevant measured period. Assume a service is measured continuously over a thirty-day month of 720 hours and is unavailable for 3.6 hours, with no exclusions. Availability is 716.4 divided by 720, or 99.5%. A percentage is meaningful only with its measurement basis: excluding maintenance windows, measuring only branch hours or ignoring a failed channel can produce a different figure without improving the customer’s actual experience.
- In the stated continuous-month example, 3.6 hours downtime gives 99.5% availability.
- Measurement scope and exclusions affect the interpretation of availability statistics.
Capacity planning examines peak demand and expected growth across processing, storage, networks and supporting infrastructure. Average daily volume can conceal salary-day, month-end or payment-event peaks. A system that handles normal traffic comfortably may struggle when many customers act simultaneously. Stress and load testing should therefore reflect realistic transactions and dependencies. Buying more storage will not necessarily fix a bottleneck caused by application locking or a slow external service; the limiting resource must be identified.
- Peak demand and dependencies matter alongside average usage.
- Capacity remedies should address the actual bottleneck rather than an arbitrary component.
Migration moves data or processing from one system to another. A successful technical load does not prove a successful banking migration: account counts, balances, customer links, interest parameters, mandates and transaction history must remain correct. Reconciliation should include both aggregate totals and meaningful record-level checks. Two offsetting errors can leave the total unchanged while affecting individual customers. An approved cutover plan should also specify responsibilities, exception resolution and what happens if the new environment cannot safely operate.
- Migration validation should test both totals and individual record relationships.
- Equal aggregate balances can conceal offsetting customer-level errors.
For a dated operational update, RBI issued its Commercial Banks-Cybersecurity, Technology: Risk, Resilience and Assurance Framework on 31 July 2026 with immediate effect; the reviewed text is updated through 1 October 2026. Its defined commercial-bank scope excludes small finance banks, payments banks and local area banks. The reviewed provisions address migration integrity and sign-offs, environmental protection, capacity assessment and network controls. These post-June developments must be distinguished from an examination session using an earlier cutoff; the central project cutoff note explains that distinction.
- The reviewed operational framework is dated July 2026 and updated through October 2026.
- Institutional scope and examination cutoff must be checked separately.
9. Data warehouses and management analysis
Operational processing answers questions such as the balance available now or whether a repayment has posted. A data warehouse organises information from relevant sources for analysis across subjects and time. It may combine account, transaction and other authorised business data after integration and quality checks. A warehouse is not simply a backup copy of the core database: its design serves reporting and analysis. Its information may have a loading delay, so a historical analytical balance should not automatically be used to authorise a current withdrawal.
- Operational systems support current transactions; warehouses support integrated analysis over time.
- A warehouse is neither automatically a backup nor the current withdrawal-authorisation record.
Extraction, transformation and loading describe a common movement of data from sources into an analytical environment. Transformation can standardise dates, units, identifiers and classifications, resolve defined inconsistencies and derive agreed measures. The process should preserve lineage so an analyst can trace a result to its origin and transformation rules. If one source stores amounts in rupees and another in thousands of rupees, combining them without conversion can create a thousandfold error even when every file loads successfully.
- ETL extracts, transforms and loads data for the target environment.
- Consistent units and traceable transformations are essential to reliable analysis.
Online transaction processing, or OLTP, commonly handles many relatively focused operational transactions. Online analytical processing, or OLAP, supports analysis such as comparing loan growth by product, region and period. These are workload distinctions, not a claim that modern technology cannot support both in a broader platform. A data mart focuses on a particular business area. Whatever the architecture, the bank should use consistent definitions so that two departments do not describe different populations as the same “active customer” measure.
- OLTP and OLAP describe different processing purposes and workloads.
- A data mart has a focused analytical scope; shared definitions support comparable reports.
10. Data mining, interpretation and revision case
Data mining examines data for useful patterns or relationships, using methods that may include classification, clustering, association and anomaly detection. Classification assigns cases to defined categories; clustering groups similar observations without those predefined labels. An anomaly may merit investigation, but unusual does not automatically mean fraudulent. A pensioner’s first large medical payment can look different from past behaviour for a legitimate reason. Analytical alerts should support informed review and proportionate action rather than replace judgment with a label.
- Classification uses defined categories; clustering discovers similarity-based groups.
- An anomaly is a reason for investigation, not proof of wrongdoing.
Suppose a model flags 100 transactions and investigation confirms fraud in 20. In this simplified labelled set, precision is 20%. That number does not reveal how many fraudulent transactions the system missed; recall requires the total number of actual frauds in the relevant population. For example, if there were 40 actual frauds and 20 were detected, recall is 50%. These measures expose different weaknesses and should be interpreted alongside operational cost, customer inconvenience and the quality of the underlying labels.
- Example precision is 20/100 = 20%; example recall is 20/40 = 50%.
- Precision and recall require different denominators and measure different aspects of detection.
A branch reports that enquiries are slow, balances differ from a monthly dashboard and a terminal shuts down during a power cut. These observations need separate investigation: network or processing delay, analytical refresh timing and power support respectively. Increasing bandwidth cannot automatically fix a stale warehouse or an exhausted UPS battery. The officer should identify the affected service, preserve transaction references, follow escalation procedures and communicate verified facts to customers. Computerisation is most useful when staff understand the system well enough to recognise its limits.
- Diagnose performance, data freshness and power continuity as distinct issues.
- Preserve evidence and use approved escalation rather than guessing transaction outcomes.
UPS runtime illustration
Assume usable stored energy after losses and a constant connected load. This is a teaching estimate.
Chapter practice
This practice is separate from timed mock tests. Answers and explanations appear after you choose an option.
Your reading progress
Sign in to save your progress across devices.