Post-Go-Live Oracle ERP Cloud Controls
The controls approved at go-live do not stay frozen. People change jobs, teams copy Job Roles, new integrations are introduced, and exceptions meant to be temporary can become part of the process. The question is not whether your Oracle Fusion Cloud ERP environment has changed, but whether those changes were reviewed, and whether the controls you rely on still work.
How has your control environment changed since go-live, and what needs action now? Compare current access, role design, configuration, transactions, and evidence with the approved go-live baseline. Investigate differences in their business context before calling them failures. Then prioritize confirmed issues by impact, affected population, recurrence, and reliance on the control. The outcome should be a recovery plan with owners, deadlines, and retests.
When is a health check useful?
The first review often makes sense after hypercare, once teams have worked in production long enough for access requests, exceptions, and workarounds to emerge. Run another before a significant audit or controls-testing cycle, after a reorganization or major process change, or when repeated findings suggest an issue was never resolved at its source. A quarterly update can also prompt a focused review of affected security settings, workflows, reports, and integrations.
Set the boundaries before pulling data. Specify the review period, in-scope financial processes, Business Units, Ledgers, Legal Entities, Inventory Organizations, integrations, and controls. A broad export of users and Job Roles may reveal differences, but it cannot tell you which ones matter without an agreed scope.
What should you compare against?
Start with the approved go-live design, not an idealized version of how the system should work today. Gather the approved Job Roles, user assignments, Data Roles, and Data Access; the segregation-of-duties (SoD) rules and sensitive Privileges used for risk decisions; key configuration and workflows; the control inventory; test results; accepted exceptions; and named business, role, and control owners.
For each area, establish three things: what was approved, what exists now, and why it changed. A new Business Unit or integration may explain a difference. It does not remove the need to check whether the change was authorized, whether its control impact was assessed, and whether it operates as intended.
| Area | Baseline-to-current-state question |
|---|---|
| Access | Which users gained, retained, or lost Job Roles and Data Access and were those changes approved? |
| Role design | Have copied Job Roles, inherited Duty Roles, or Privileges changed effective access? |
| Configuration | Do current settings, approvals, integrations, and reports match the approved design or later authorized changes? |
| Transactions | Are critical activities following the expected approval and exception-handling paths? |
| Evidence | Can owners show complete records for the controls and periods under review? |
Treat the differences as leads. A change is not, by itself, a control failure.
Has access drifted beyond the approved design?
Reconcile active Fusion users and their Job Role and Data Access assignments with worker records and approved requests. Focus first on access that should have changed: leavers whose accounts or assignments remain active, movers who kept access from a previous position, temporary elevated access without an end point, and integration users or service accounts without a clear owner.
Then follow the provisioning path. Was the assignment requested and approved? Was its SoD risk assessed before provisioning? If access was removed, could a provisioning rule assign it again? Oracle’s role-provisioning guidance explains the distinction between automatic and manual provisioning—and why manually provisioned roles may need explicit removal when a worker changes roles. Correcting an individual assignment may not fix the rule or process that caused it.
Do not judge risk by a Job Role name alone. A Job Role can inherit Duty Roles and Privileges, while Data Access and security context affect which organizations and transactions a user can reach. Oracle’s User and Role Access Audit Report provides details of granted function and data security privileges. Use those details and the Security Console’s role-hierarchy view to establish effective access before escalating a finding.
Consider a user flagged for both supplier maintenance and payment processing. The useful question is not simply whether two Job Roles appear on their record. Determine which relevant Privileges they inherit, whether they can exercise them within the same business context, whether the access was approved, and whether a mitigating control operates. That analysis may confirm a conflict, narrow its scope, or show that the initial flag needs no corrective action.
Also look beyond individual users. If teams repeatedly copy Job Roles to meet small variations in access needs, the resulting growth can make ownership and certification harder. The related Oracle Cloud ERP license-controls guide examines another consequence worth checking: whether role and Privilege assignments affect subscription costs. Do not assume a specific cost impact without checking the organization’s entitlements and contract.
Have changes affected control operation?
Review the production changes made since go-live to approval rules, financial controls, reports, integrations, and other configuration that critical processes depend on. For each material change, look for the request, approval, impact assessment, test result, and production sign-off.
Keep authorization and effectiveness separate. An approved configuration change may still affect a dependent control that nobody retested. Equally, missing test evidence does not prove the control failed; it means the team cannot yet demonstrate the result. Record those as different findings. The Oracle Cloud ERP configuration-change governance guide provides a deeper treatment of routing changes to owners and recording their decisions.
Treat each quarterly update as a governance checkpoint, not as an assumed source of new risk. Relevant delivered security changes, enabled functionality, and customer role configurations may call for reassessment. Oracle says it applies Fusion Applications updates to Test before Production, creating a window to check affected processes. Functional regression tests should be supplemented by risk-based checks of the access and controls those processes rely on.
For example, if an approval workflow changes, test more than whether a transaction can still be submitted. Confirm that the intended approver receives it, a user without approval authority cannot bypass the route, exceptions are visible, and the report used to review them still returns the expected population.
What do transactions and exceptions reveal?
A control can be appropriately designed but operate inconsistently. Test a defined sample of activity in the financial processes that matter most—such as supplier maintenance, payments, journals, or bank-account changes—and trace it through the expected approval, monitoring, and exception process.
Look for repeated overrides, sensitive master-data changes, transactions outside the normal approval path, and exceptions carried from one period into the next. Discuss the circumstances with process owners. An urgent transaction with documented authorization and follow-up is different from a recurring workaround with no owner or resolution date.
Some conclusions require more than Fusion data. HR records may establish a user’s status; an identity workflow or ITSM ticket may hold the approval; a connected application may complete a step in the process. Identify which tests depend on Oracle extracts, adjacent-system records, or interviews. An interview can explain an exception, but it should not quietly stand in for evidence of a required control activity.
Trends help expose issues that a single sample misses. Track recurring exceptions, time to remediation, overdue items, and the proportion of reviewed cases with complete approval and closure evidence. Define each measure consistently so that a change between periods means something.
Is the evidence complete enough to use?
A control may have operated, yet still be difficult to substantiate if its records are fragmented. For each key control, establish what evidence should exist, who prepares and reviews it, where it is stored, and how it links to the population and period tested.
An access decision might need a request, approval, Job Role and Data Access provisioning record, SoD result, certification decision, mitigation, and remediation history. Those pieces may sit across Fusion extracts, identity workflows, ITSM tickets, spreadsheets, and shared folders. What matters is whether the record is complete, accurate, traceable, and retrievable. For a closer look at the distinction between a finding and proof of control operation, see what auditors need Oracle control evidence to show.
Check for missing populations, dates, reviewer decisions, exception dispositions, and proof of retest. If an access certification flags a conflict but never records what happened next, the team cannot demonstrate closure. If a ticket says a change was tested but contains no result, the test may need to be repeated.
Understand the limits of Oracle audit reports as well. Business-object auditing must be configured for the relevant objects and attributes; Audit Reports show history for objects enabled for auditing. The absence of an event in a report does not, on its own, establish that no change occurred. Check the audit configuration and other available records before drawing that conclusion.
Throughout the review, distinguish design effectiveness from operating effectiveness. A sensible review procedure may not have been performed. A completed sign-off may cover the wrong population. The health check should say which problem it found, rather than treating all evidence gaps as proof of the same failure.
Which findings should you fix first?
Prioritize only after you understand what each finding means. Record its business impact, how readily the access or weakness could be exercised, the number of users or transactions affected, whether it recurs, and how much the organization relies on the control. Note remediation effort separately. A simple 1-to-3 score for each risk factor can help teams compare findings, but a confirmed high-impact exposure should not wait because its total score is lower than another issue’s.
| Finding | Question to resolve | Initial response |
|---|---|---|
| A leaver’s Fusion account appears active | Can the person still sign in, and what Job Roles, Privileges, and Data Access remain? | Restrict access if confirmed; investigate the deprovisioning failure. |
| A copied Job Role includes a sensitive Privilege | Who has the role, what do they inherit, and where can they use it? | Validate the impact; correct the role or assignments and retest. |
| An approval exception recurs monthly | Is the underlying workflow, staffing, or policy creating the repeat exception? | Assign a process owner and address the root cause. |
| An access certification lacks a decision | Was the review completed, and is there reliable evidence elsewhere? | Recover the record if possible; correct the review process. |
Put the decisions in a findings register. Each entry should capture the baseline expectation, observed difference, evidence, severity and rationale, business impact, accountable owner, due date, any mitigating control, corrective action, and retest status.
For example: “HC-04: Job Role and Data Access retained after transfer; two users affected; payment-related effective Privileges under review; owner: ERP security lead; interim action: restrict access pending validation; retest: open.” That is specific enough to act on without presenting an unverified assignment as a proven loss event.
What should the recovery plan include?
Agree the top five remediation priorities and assign one accountable owner to each. Use a 30/60/90-day plan to organize the work, while containing any immediate exposure sooner.
- First 30 days: Validate the highest-risk findings, restrict confirmed inappropriate access, assign owners, document temporary mitigating controls, and resolve straightforward approval or evidence gaps.
- By 60 days: Correct the provisioning rules, copied Job Roles, Data Access assignments, configuration, or operating procedures behind recurring findings. Formally resolve or accept exceptions through the appropriate governance process.
- By 90 days: Retest corrections against defined populations and periods. Confirm that access stays corrected, controls operate, evidence is retrievable, and recurring exceptions are declining.
The plan should also establish what will be monitored between health checks. High-risk Privileges, broad Data Access, and unresolved exceptions may warrant closer attention than stable, lower-risk settings. Align update-related assessments with the release calendar and event-driven reviews with reorganizations, new integrations, and major process changes. Control owners, internal audit, and the Oracle applications team should agree the actual cadence and thresholds.
A completed health check leaves the organization with three usable outputs: a baseline-to-current-state comparison, a findings register that distinguishes suspected from confirmed issues, and an owned plan to correct and retest the priorities. The Oracle Cloud ERP risk-to-resolution guide goes deeper on moving from investigation through decision, implementation, and documented closure.
Common follow-up questions
Is every change since go-live a control failure?
A change may reflect an approved business need, an accepted exception, or incomplete documentation. Validate effective Privileges, Data Access, process context, approval, and operating evidence before classifying it.
How is this different from an access review?
An access review asks whether users still need their assignments. A post-go-live health check also tests role inheritance, provisioning, configuration, transactions, exceptions, update impacts, and evidence that controls operated during the period.
What if the go-live baseline is incomplete?
Use the best available approved artifacts and state what cannot be reconstructed. Ask accountable business and technical owners to validate the intended control state, then document the resulting baseline and its limits. Do not use a newly agreed baseline as proof that the same control operated in earlier periods.
Next step: Run a scoped comparison against your approved go-live baseline and agree your top five remediation priorities. Read How Oracle ERP Cloud Access Puts Your Next Audit at Risk
See governance applied to the access you have today
A working session with a governance specialist — not a slide presentation.
Book your tailored demo