Is restoring a backup preventive, detective or corrective? Does a passed range check prove the data is right? This lesson covers CISA Domain 1, Section 1.4: what a control is, how to classify it, and what an auditor checks before relying on it. Watch the full lesson above, or read the notes below.
1.4.1 Internal controls
Internal controls are the policies, procedures, practices and organisational arrangements that give reasonable assurance that objectives will be achieved. They:
- Protect assets such as money, data and equipment
- Support reliable information that people can trust
- Promote effective operations
- Meet compliance obligations
Management designs and operates controls. The auditor evaluates whether their design and operation address the risk.
1.4.2 Control objectives and control measures
Think of a control in three levels, using one example:
| Level | Meaning | Example |
|---|---|---|
| Control objective | The desired result | Prevent unauthorised supplier changes |
| Control | The measure | Approval by a second, separate person |
| Control practice | How it's carried out | Compare a verified bank instruction with the requested change before release |
Remember that a procedure can exist on paper while the control isn't operating.
How it's implemented vs what it's for
These are two separate questions:
- Implementation: administrative controls set direction (policies, training), technical controls enforce or record activity in systems, and physical controls protect premises and equipment.
- Purpose: preventive, detective or corrective.
So a technical control can be preventive or detective.
IT general controls vs application controls
| IT general controls (ITGCs) | Application controls |
|---|---|
| Access administration, change control, operations, physical protection | Specific transaction processing |
| Support the whole environment | Work at input, processing and output |
Take an invoice application. It rejects an unknown supplier at input, calculates the amount during processing, and restricts who gets the payment report at output. Those are application controls. But if an administrator can change the program without approval, weak general controls undermine all three.
Exam tip: before relying on an automated control, test the general controls it depends on, especially change management and privileged access.
Input checks the exam loves
| Check | What it does | Example |
|---|---|---|
| Validity | Confirms a value exists in an approved set | Supplier ID is in the master file |
| Check digit | A calculated extra digit | Catches a mistyped identifier |
| Sequence | Finds gaps | A missing invoice number |
| Duplicate | Flags repeats | Same invoice entered twice |
| Completeness | Confirms required fields are present | No blank mandatory fields |
| Range / limit | Value lies between bounds | Month from 1 to 12 |
| Reasonableness | Value is plausible in context | Overtime against hours worked |
| Exception report | Lists items outside the rules | Items for investigation |
Each check catches only its own error. A month of 18 fails the range check. A month of 8 passes it, even if the invoice belongs in September. Passing one check never proves every part of a record is correct.
Control totals
- Batch total: sums a meaningful amount, like invoice value.
- Hash total: sums a numeric field used only as a check, like account numbers. It is not a cryptographic hash.
- Record count: checks the number of records.
- Run-to-run totals: reconcile records accepted, rejected and posted across processing stages.
Matching totals can hide offsetting errors. If two records are lost and two duplicates are added, the record count still matches. A financial total, an identifier comparison or a transaction-level reconciliation could reveal what the count alone misses.
Risk shapes the controls
Different applications need a different focus:
- E-commerce: internet-facing inputs can enable injection, so check secure design and input validation.
- EDI: partner identity, message integrity and duplicate handling.
- Industrial control systems (ICS/SCADA): safety, segmentation and controlled remote access.
- AI and expert systems: validation, bias and human oversight.
1.4.3 Control classifications
Design vs operating effectiveness
| Design effectiveness | Operating effectiveness |
|---|---|
| Could the control address the risk if performed properly? | Did it actually work, consistently, during the period? |
| Payroll example: the business owner approves access before provisioning | Approvals exist, they're genuine, and they predate the access |
Five control purposes
- Preventive: stops an unwanted event, like requiring approval before a payment.
- Detective: identifies it, like a reconciliation that spots an unauthorised payment.
- Corrective: repairs the consequences, like correcting the record and recovering funds.
- Deterrent: discourages misconduct, like visible monitoring notices.
- Compensating: addresses the objective when the preferred control is impractical, like a separate review where full segregation of duties isn't possible.
Exam tip: controls can have more than one purpose, so classify the function the question describes. A log that records an event is detective. An automated response that blocks the activity is preventive.
1.4.4 Controls and risk
A control should address an identified risk to a business objective.
- Preventive controls reduce likelihood.
- Detective and corrective controls help find and limit the consequences.
So assess the combination, and the residual exposure. Adding a control is useful only if it addresses the risk and its benefit justifies its cost.
1.4.5 Prescriptive controls and frameworks
Common references:
| Framework | Focus |
|---|---|
| CIS Critical Security Controls | Prioritised security safeguards |
| PCI DSS | Payment card data, where scope matters |
| CSA Cloud Controls Matrix (CCM) | Provider and customer responsibilities |
| COBIT | Connects business needs, IT processes and assurance |
Selecting a framework is not evidence that its controls work. Map the relevant requirements to risk, identify the evidence, and document justified exclusions. A green dashboard is only useful if the measurement behind it is complete, accurate and relevant.
1.4.6 Evaluating the control environment
The control environment is the culture, oversight and accountability that support control.
| Management monitors | Audit evaluates |
|---|---|
| Reviews exceptions and reconciles activity | Corroborates self-reported results through records, observation and testing |
| Tracks corrective action | Looks closely at what happens when controls fail |
A report produced but never reviewed isn't monitoring. A well-written policy can't make up for routine override. Every control is limited by error, collusion, override and cost, so controls give reasonable assurance, never absolute.
Exam traps
- Calling a backup restore preventive. It acts after the event, so it's corrective.
- Assuming a passed validation check proves the record is right. Each check catches one kind of error.
- Relying on an automated control without testing the general controls behind it.
- Treating a chosen framework as evidence that controls work. Selecting isn't testing.
Practice questions
Question 1. After a storage failure, operations staff restore a customer database from the previous night's backup. Which classification BEST describes the restoration?
A) Preventive, because it limits the loss of customer data B) Detective, because it reveals the corruption to operators C) Corrective, because it recovers from an event's effects D) Compensating, because it offsets weak storage controls
Answer: C. Restoration acts after the loss, so it isn't preventive, and detection happened before the restore. A compensating control substitutes for a missing control, while restoration is a primary recovery measure.
Question 2. During a batch transfer, two payment records are lost and two other records are duplicated. Which control is LEAST likely to detect the problem?
A) A comparison of record counts before and after transfer B) A batch total of payment amounts before and after transfer C) A comparison of transaction identifiers across both files D) A transaction-level reconciliation of the two files
Answer: A. Two lost plus two added leaves the count unchanged. A batch total would usually change unless the values happen to offset, and identifier comparisons and reconciliations match record by record.
Question 3. An IS auditor plans to rely on an automated three-way match in the payables application. Which area is MOST important to test before placing that reliance?
A) The accuracy of last month's payables exception report B) Controls over changes to the application program C) The completeness of supplier master data fields D) Management's approval of the invoice payment policy
Answer: B. An exception report shows one month's output but says nothing about whether the matching logic stayed unchanged. Master data is an input concern, and an approved policy shows intent, not operation.
Keep going
- Read the free companion book, the CISA Complete Study Guide (First Edition 2026) by Muhammad Aamir.
- Watch the full CISA Complete Course by Mirak X. Next up is Section 1.5, Audit Project Management.
- Subscribe to Mirak X on YouTube. New lesson every day at 6 PM (PKT).
- Try free CISA practice questions and mock exams at https://certprepx.com/.
Free study material. Not affiliated with or endorsed by ISACA. CISA is a registered trademark of ISACA.


