What should a CPS 230 critical operation map actually show?
A critical operation map is a set of linked artefacts rather than one diagram: the end-to-end process with its steps and hand-offs, and for each step the people and roles, technology, information, facilities and service providers it depends on, with the tolerance level the operation must stay within. That is what CPS 230 paragraphs 26(b) and 37 ask you to identify and document; the words "process map" appear nowhere in the standard, so the form is yours to choose and the content is not.
Written for operational risk, resilience and process teams at APRA-regulated entities who have a list of critical operations and now need the maps behind them. Reviewed by Mike Bentley, founder, .
Linked artefacts, not one drawing
A critical operation map answers one question: if this resource fails, which steps stop, and does the operation stay within tolerance? One diagram cannot, because it carries sequence and hand-offs and nothing else. The map that works is a set of linked artefacts: the end-to-end process as steps with hand-offs, and attached to each step the people and roles who perform it, the technology it runs on, the information it needs, the facility it happens in and the provider it relies on, with the tolerance level the operation must stay within.
That shape follows the standard's own list. Prudential Standard CPS 230 Operational Risk Management, paragraph 26(b), requires an entity to "identify and document the processes and resources needed to deliver critical operations, including people, technology, information, facilities and service providers, the interdependencies across them, and the associated risks, obligations, key data and controls". Paragraph 34 defines critical operations as "processes undertaken by an APRA-regulated entity or its service provider" whose disruption beyond tolerance levels would have a material adverse impact. A critical operation is a process, so its map starts as a process map and grows dimensions a process map alone does not carry.
Our earlier article on what CPS 230 requires you to document covers why this is a mapping problem, not a register problem; this one is about what the map should show. It is a plain summary for the people doing the work, not prudential or legal advice.
What APRA requires, and what is sound practice
Paragraph numbers here follow the CPS 230 and CPG 230 texts dated July 2026: the amended standard APRA determined on 23 April 2026, which its paragraph 6 says commenced on 1 July 2026, and the practice guide APRA revised with it. The earlier article covers how the standard got here; this one is about the map.
The obligations that shape the map: paragraph 26(b), identify and document the processes and resources for each critical operation; paragraph 33(a), a register of critical operations; paragraph 37, tolerance levels for each; paragraph 39, a business continuity plan holding that register, its tolerance levels and the "key internal and external dependencies"; paragraph 48, a register of material service providers.
What it does not say matters as much. Neither CPS 230 nor CPG 230 uses the words "process map" or "mapping", in the June 2024 guide or the July 2026 revision (full text checked as at 9 November 2026). CPG 230, which says practice guides "do not themselves create enforceable requirements", adds the sound practice: senior management "should be satisfied that they have sufficient detail about the resources and processes needed to deliver critical operations", and prudent entities would keep documented processes in the framework and up to date (paragraphs 19 and 20).
Five resource dimensions, captured per step
A resource recorded against the operation says the operation uses it. Recorded against a step, it says where the operation stops when that resource is unavailable, which is what a tolerance level is measured against. So each dimension is captured per step, by reference to the register that owns the detail.
Paragraph 26(b) also lists interdependencies, risks, obligations, key data and controls. On a step-level map those are attributes of the step: the control operating there, the obligation discharged, the risk recorded.
- People and roles. The role that performs the step and the role that approves it, as references to the HR or identity source, plus how many people can do the step today.
- Technology. The applications and infrastructure the step runs on, including the integration between two systems and the scheduled job nobody "uses", referenced to the technology asset register so each recovery time is read, not guessed.
- Information. The data the step needs present and correct, where it comes from, and its place among the information assets CPS 230 paragraph 24 asks you to monitor.
- Facilities. Where the step is physically performed and whether it can be performed elsewhere: the data centre as well as the office, and a provider's premises when the step runs there.
- Service providers. The provider that performs or supports the step, as a reference to its entry in the material service provider register, with the fourth parties it relies on where known.
Linking providers so register and maps agree
CPS 230 paragraph 48 requires a register of material service providers; paragraph 50 requires its annual submission to APRA. CPG 230 paragraph 48 says a prudent entity's register would list each material arrangement with a responsible person, "which critical operation(s) the material arrangement supports" and, where the arrangement delivers a critical operation, the fourth parties involved. Paragraph 31 of the guide adds that the register of critical operations would typically include the provider arrangements supporting each. The two registers already point at each other.
The map is where that pointer becomes specific. A register can say a provider supports payments; only the map says which step, what happens there when the provider fails, and whether the alternative keeps the operation inside tolerance. The link is made at the step, with the provider referenced from the register, not retyped.
Two reconciliations keep the artefacts honest. Every provider on a step of a critical operation's map is in the register or has a documented reason why not. For a provider APRA prescribes as material, CPG 230 paragraph 49 expects that reason to be approved by an Accountable person under the Financial Accountability Regime and reviewed at least annually, and the same treatment is a sensible default for any other. Every register entry linked to a critical operation appears on a step of that operation's map. The April 2026 amendments exempt arrangements with the provider categories in the attachment to CPS 230 from specific agreement, exit and monitoring requirements where the arrangement uses standardised terms or has no formal agreement (paragraph 57), but CPG 230 Table 5 still expects them included "when identifying processes and resources needed to deliver critical operations" and on the register.
A tolerance level attached to a process
A tolerance level is the limit of disruption an entity decides it will accept for a critical operation. CPS 230 paragraph 37 requires three for each: "the maximum period of time the entity would tolerate a disruption", "the maximum extent of data loss the entity would accept" and "minimum service levels the entity would maintain while operating under alternative arrangements". CPG 230 Table 4 translates these into recovery time, recovery point and recovery level objectives, and says an entity "would normally determine the minimum level of people, information assets and other resources required to provide the business service".
Attached to a document, a tolerance is a number in a table. Attached to a process, it becomes a test each step is checked against: compare each step's recovery time, people depth and provider fallback with the operation's maximum period. Where a step's own recovery is slower than its operation's tolerance, the map has found a gap. The minimum service level needs the map most, because the minimum level of people, information assets and other resources is the step-level resource list with non-essential steps switched off.
This also makes two obligations answerable. Paragraph 40 requires the entity to monitor its performance against tolerance levels and report any failure to the Board; paragraph 41 requires APRA to be notified no later than 24 hours after a disruption to a critical operation outside tolerance. You only know an operation is outside tolerance if the clock runs against the operation, not the system that failed.
Building the maps when documentation is thin
Most critical operations run partly in procedures and partly in people's heads, and the sequence in the procedure is rarely the one performed. The sources that exist each supply one dimension: contracts and service schedules the providers, the technology asset register the systems, the organisation chart and rosters the people, the information asset register the data, the facilities list the places. None supplies the steps.
The steps come from the people who do the work: interviews, meeting transcripts, the wiki page someone wrote for a new starter, and a recording of the job being done once, which gives the real sequence and the applications touched. Our article on mapping a process when nothing is documented sets out that method; it applies unchanged here. Treat every drafted step as unverified until the person who performs it confirms it; a map that looks complete and is wrong fails when consulted.
Omni by Valstra is one way to hold the result: processes, people, systems, data objects and documents in one model, generated from the documents, transcripts and wiki pages the entity already has, with a consented recording of the job done once supplying the steps, and every change an AI proposes accepted row by row. A spreadsheet per operation with a reference column into each register is another, and for a small entity with a short list a reasonable one. The test for either: pick a resource and see which steps stop.
Keeping the maps current through change
CPS 230 paragraph 44 requires the business continuity plan to be updated, as necessary, annually to reflect changes in legal or organisational structure, business mix, strategy or risk profile, and CPG 230 paragraph 40 adds "as soon as possible after a material change in the entity's structure, business or risk profile". Paragraph 60(a) requires APRA to be notified within 20 business days of entering into or materially changing an agreement for a service the entity relies on for a critical operation. A map maintained only at the annual review is wrong for most of the year.
So attach each dimension of the map to the event that changes it. A new or amended provider agreement updates the steps that provider serves. A system decommissioning updates the technology on its steps. A restructure updates the roles. A changed tolerance re-runs the step check above. CPG 230 paragraph 43 names "changes to services, processes, information assets, the business environment and stakeholder expectations" as triggers for additional assurance; the same list triggers the map.
Give each map an owner, as CPG 230 paragraph 48 gives each arrangement a responsible person, and keep the superseded version when a map changes, so what was true at the time of a disruption can be shown. Align the per-operation review with the annual update of the business continuity plan and the annual provider register submission under paragraph 50, so the three artefacts are reviewed as one set.
Sources
- APRA, Prudential Standard CPS 230 Operational Risk Management (July 2026)
- APRA, Prudential Practice Guide CPG 230 Operational Risk Management (July 2026)
- APRA, Operational risk management: CPS 230 and CPG 230 updates and the material service provider register template
- APRA, CPS 230 marked-up version showing the April 2026 amendments (PDF)
Common questions
- Is a diagram enough?
- Not on its own. A swimlane shows steps and hand-offs, which is part of what CPS 230 paragraph 26(b) asks for, but it does not carry the technology, information, facilities, providers, obligations, key data and controls per step, or the tolerance level. Keep the diagram and attach the rest to its steps.
- What if a step's recovery time is longer than the operation's tolerance?
- The map has found a gap, which is its job. Either the step's recovery is brought inside the tolerance, an alternative arrangement that keeps the operation within tolerance is written into the business continuity plan CPS 230 paragraph 39 requires, or the tolerance itself is revisited and the change recorded; the map is not edited to hide it.
- Do we need one map per critical operation or per process?
- CPS 230 paragraph 34 defines critical operations as processes, so they are often the same thing. Where one critical operation, such as payments, is delivered by several distinct processes, map each process and record which critical operation it belongs to, so the register of critical operations rolls up from the maps rather than being drawn separately.
- How do the maps relate to the material service provider register?
- The register, required by CPS 230 paragraph 48 and submitted annually under paragraph 50, lists the providers and, as CPG 230 paragraph 48 suggests, which critical operations each arrangement supports. The map shows the step where that support lands. Reconcile both ways: every provider on a map is in the register or has a documented reason, and every register entry linked to an operation appears on a step of its map.
See it against your own processes
A demo runs on your material, not a canned dataset. Bring a process you find hard to explain.