Skip to content

CISA Domain 3 - IS Acquisition, Development, and Implementation (12%)

Project governance, the development lifecycle, and the controls that must exist before a system goes live. The auditor's angle is whether controls were built in, not bolted on.

Project governance

  • Business case - the justification: expected benefits, costs, risks, and alternatives. Approved before funding.
  • Feasibility study - assesses whether the proposal is achievable technically, economically, operationally, and legally.
  • Project sponsor - the business owner accountable for benefits. A project without an engaged sponsor is a standard finding.
  • Project steering committee - approves scope, budget, and major changes.
  • Critical path - the longest sequence of dependent tasks; it determines the minimum project duration. Slippage on the critical path slips the project.
  • Function point analysis - sizing software by delivered functionality rather than lines of code, used for effort estimation.
  • Earned value analysis - compares planned versus actual cost and schedule progress.

The auditor's involvement should begin early. Reviewing controls only at go-live is too late to influence design, which is a recurring exam theme.

Development methodologies

  • Waterfall - sequential phases with formal sign-off between them. Suits stable, well-understood requirements; expensive to change late.
  • Agile - iterative delivery in short increments with continuous stakeholder involvement. Suits evolving requirements. Documentation is lighter, which is an audit consideration rather than an excuse for none.
  • Prototyping - building a working model to clarify requirements. Risk: the prototype is promoted to production without proper controls.
  • RAD (Rapid Application Development) - compressed timelines with heavy user involvement.
  • DevOps - development and operations integration with automated pipelines. See CI/CD explained.
  • Reverse engineering - deriving design from an existing system, often for replacement.

SDLC phases and the controls in each

  1. Feasibility - business case and alternatives.
  2. Requirements definition - functional and control requirements. Security and audit requirements belong here, not later.
  3. Design - architecture, data model, interfaces, and control design.
  4. Development - coding against standards, with code review and version control.
  5. Testing - unit, integration, system, and acceptance.
  6. Implementation - conversion, go-live, and handover.
  7. Post-implementation review - did it deliver the expected benefits?

  8. User acceptance testing (UAT) - performed by business users against business requirements. Users, not developers, must sign off.

  9. Regression testing - confirms existing functionality still works after a change.
  10. Parallel testing - old and new systems run together and results are compared.
  11. Pilot testing - deployment to a limited group before full rollout.
  12. Test data - must be representative. Using live production data in test raises a privacy exposure; it should be masked or anonymized.

Implementation and conversion

  • Direct (big bang) cutover - old system off, new system on. Cheapest, riskiest, no fallback.
  • Parallel changeover - both systems run; highest cost, lowest risk, allows result comparison.
  • Phased changeover - implemented in stages by module or location.
  • Pilot changeover - one site or group first.
  • Data conversion controls - record counts, control totals, and reconciliation before and after migration. Conversion is where data integrity is most often lost.
  • Fallback / rollback plan - how to return to the previous state. Required before go-live.
  • Post-implementation review (PIR) - conducted after the system has stabilized, comparing actual benefits to the business case.

Parallel running is the lowest-risk conversion and direct cutover the highest. Questions about a critical system with no tolerance for failure point to parallel.

Application controls

  • Input controls - validation, edit checks, range checks, check digits, sequence checks, and batch totals. Preventing bad data entry is cheaper than correcting it later.
  • Processing controls - run-to-run totals, reasonableness checks, and reconciliation.
  • Output controls - distribution restrictions, reconciliation of totals, and error reporting.
  • Check digit - a calculated digit appended to a number to detect transcription errors.
  • Hash total - a meaningless-but-verifiable sum used to confirm nothing changed in transit.
  • Audit trail - a record of transactions sufficient to reconstruct what happened and who did it.

Change and release management

  • Change management - request, assess, approve, test, implement, review. Emergency changes still need retrospective approval and documentation.
  • Configuration management - controlling and recording versions of code and configuration.
  • Version control - the authoritative history of source code.
  • Library control - separation of development, test, and production libraries.
  • Migration to production - performed by someone independent of the developer. Developers with production migration rights is one of the most common CISA findings.
  • Emergency change (firecall) access - temporary elevated access, logged, time-limited, and reviewed afterwards.

Acquisition of third-party systems

  • Request for proposal (RFP) - documented requirements issued to vendors for comparable bids.
  • Vendor evaluation criteria - defined and weighted before proposals arrive, to keep selection objective.
  • Escrow agreement - source code held by a third party and released if the vendor fails. The control for vendor viability risk in critical bespoke software.
  • Contractual right to audit - retained access to assess the vendor's controls.
  • Acceptance criteria - objective tests the delivered system must pass before payment or go-live.

Exam pointers

  • Security and control requirements must be defined in the requirements phase. Retrofitting is the wrong answer.
  • UAT is signed off by users. Developer sign-off is a finding.
  • Production data used in testing without masking is a privacy exposure.
  • The developer must not migrate their own code to production. Look for segregation of duties.
  • Source code escrow addresses vendor failure risk specifically, not quality or support.
  • A post-implementation review measures benefits against the business case, not technical success alone.

Official documentation

πŸ“– ISACA CISA exam content outline - authoritative domain list πŸ“– COBIT framework - build, acquire, and implement objectives