An identity governance platform only governs the applications that have been onboarded to it, and in most programmes that is a minority of the estate. The rest sit outside reviews, joiner-mover-leaver automation and separation of duties checks, which is exactly where orphaned accounts and forgotten admin rights accumulate. This guide covers the practical work of bringing applications under governance: prioritising the backlog, building an entitlement catalogue that reviewers can actually understand, choosing an integration pattern for connected and disconnected apps, matching accounts to identities, and running the baseline review. It is for IGA programme leads, identity engineers and application owners, and it assumes the platform has already been selected (see the IGA Buyer's Guide).
Key takeaways
- Onboarding is where IGA programmes stall. Platforms are usually live within months, but application coverage grows slowly because every app needs an owner, entitlement data and an integration decision.
- The entitlement catalogue is the product. Reviews, access requests and SoD rules are only as good as the descriptions, owners and risk ratings attached to each entitlement.
- Disconnected apps still count. An application with no connector can be governed with scheduled account imports and ticket-based fulfilment; leaving it out entirely is the real failure.
- Correlation decides data quality. Accounts that cannot be matched to an identity are either orphans, non-human accounts or bad data, and each needs a different outcome.
- Baseline first, automate second. Review existing access before you open self-service requests, otherwise automation preserves years of over-provisioning.
Why application onboarding stalls
A typical enterprise runs hundreds of applications, and the IGA platform starts with a handful: the directory, HR, perhaps the ERP. Each additional app competes for scarce engineering and application owner time, and the work is less about connectors than about information nobody has written down: who owns this app, what each entitlement means, which ones are privileged, and which accounts belong to people who have left.
- No accountable owner: application owners see governance as an audit chore and do not supply entitlement descriptions or review decisions.
- Technical entitlement names: a reviewer shown
GRP_FIN_AP_RW_02cannot judge whether it is appropriate, so they approve it. IDPro's article on optimising recertifications identifies this lack of context, and the review fatigue it causes, as the root of rubber-stamping. - Connector-first thinking: teams wait for a full read-write connector before onboarding, when read-only aggregation would already enable reviews and orphan detection.
- Dirty account data: shared, generic, test and service accounts mixed with human accounts, with inconsistent identifiers, make correlation slow and contentious.
Prioritising the onboarding backlog
Start from an application inventory (CMDB, SSO catalogue, SaaS discovery, finance records of software spend) and score each application. A simple, defensible model uses four factors.
| Factor | Higher priority when | Evidence source |
|---|---|---|
| Data sensitivity | Holds regulated, financial or customer data, or controls production | Data classification, DPIA records |
| Privilege | Has admin roles that can change other users, payments or infrastructure | Application role documentation, PAM inventory |
| Audit scope | In scope for SOX, PCI DSS, ISO/IEC 27001 or sector regulation | Audit plan, control matrix |
| Population and churn | Many users, contractors or third parties; frequent movers | HR data, SSO sign-in counts |
Group applications into waves of 10 to 20 so that patterns repeat: one wave of SaaS apps with SCIM, one of database-backed internal apps, one of disconnected apps handled by file import. Waves make effort predictable and let you report coverage as a percentage of risk-weighted applications rather than a raw count.
Building the entitlement catalogue
An entitlement is any grant of access inside an application: a role, group, permission set, licence, profile or responsibility. The catalogue is the curated record of those entitlements with the business metadata that governance processes need. Microsoft's guidance for defining application governance policies, for example, has you list each application role, decide which need approval, and record prerequisites, approvers, access duration, incompatible roles and Conditional Access requirements per role. Whatever the platform, capture the same fields.
| Catalogue field | What good looks like | Used by |
|---|---|---|
| Business description | One plain sentence on what the holder can do, written by the owner, e.g. "Can approve supplier invoices up to 50,000" | Reviewers, requesters |
| Entitlement owner | A named person (not a team mailbox) with a deputy | Approvals, reviews |
| Risk rating | High, medium or low against published criteria; privileged flag set where the entitlement can administer the app or other users | Review cadence, request approval stages |
| Requestable and approval path | Whether users can request it, and the stages (manager, then owner, for example) | Access request workflow |
| Prerequisites | Employment type, department, location or training required | Policy enforcement |
| Maximum duration | Indefinite, project-bound, or fixed (30 or 90 days) for temporary and third-party access | Expiry automation |
| SoD conflicts | Explicit incompatible entitlements, inside and across apps | Preventive and detective SoD checks |
| Birthright or role mapping | Which business roles include it by default, if any | Joiner and mover automation |
- Write descriptions with the owner, not for them: a 30-minute workshop per application, working through the entitlement export, produces better text than any template sent by email.
- Publish risk criteria: for example, high means it can move money, change access for others, read regulated data in bulk or change production. Without criteria every owner rates their app as medium.
- Collapse noise: hide entitlements that nobody holds, merge technical sub-permissions that are always granted together, and keep the raw value for traceability.
- Feed roles later: a clean catalogue is the input to role mining. See the Role Mining and Role Design Guide.
Integration patterns for connected and disconnected apps
Choose the lightest pattern that delivers the governance outcome you need now, and plan to move up. Read access (aggregation) gives you visibility, reviews and orphan detection; write access (provisioning) gives you automated joiner, mover and leaver changes.
| Pattern | How it works | Governance outcome | Trade-offs |
|---|---|---|---|
| Standards-based provisioning (SCIM 2.0, RFC 7643 and 7644) | Platform creates, updates and deactivates users and groups through the app's SCIM endpoint | Full lifecycle automation and reviews with automatic removal | Many SaaS SCIM implementations do not expose all entitlements; check group and role support |
| Directory or group-based | App authorises from AD or LDAP groups; IGA governs group membership | Reviews and removal via group changes | Only as good as the group design; nested groups hide effective access |
| Direct connector (database, LDAP, REST or SOAP API, vendor connector) | Platform reads and writes the app's own user store | Lifecycle automation for internal and legacy apps | Connector build and maintenance; a privileged service account the connector uses |
| Read-only file import (disconnected) | Scheduled CSV export of accounts and entitlements, imported and correlated | Visibility, reviews and orphan detection | Removals are manual; stale if exports stop |
| Ticket-based fulfilment (disconnected) | Requests and review revocations create tickets for the app team, with closure evidence | Auditable changes without a connector | Slow; needs SLA tracking and reconciliation against the next import |
Microsoft's Entra ID Governance documentation illustrates the range on one platform: OpenID Connect, SAML, SCIM, SQL and LDAP integrations; a provisioning agent for on-premises SCIM, SQL, LDAP and web services apps; and, for apps with no provisioning support, exporting the user list to CSV and creating application role assignments so that those users appear in access reviews. It is explicit that users not represented in the governance platform are not included in reviews, and that for apps without provisioning, denied users must be removed from the app's own store by other means. That is the disconnected app problem in one sentence.
Decision guidance: a small organisation with mostly SaaS should insist on SSO plus SCIM at procurement and use file imports for the remainder. A large organisation should run a connector factory for its top internal apps and accept read-only imports for the long tail, with ticket closure checked against the next import. See the SCIM Provisioning Guide.
Account correlation and the orphan problem
Correlation links each application account to an identity in the authoritative source, usually HR for employees and a contractor or vendor system for third parties. Pick a stable, unique matching attribute (employee ID is better than email, which changes on marriage, rebranding or rehire) and run correlation as a report before you commit it.
- Match: correlate on the primary attribute, then secondary rules (email, then username) for the remainder, recording which rule matched.
- Classify the unmatched: Microsoft's guidance for disconnected apps lists the usual causes: identifiers changed in the directory, people who have left, and vendor or super-admin accounts with no matching person. Add shared accounts, test accounts and service accounts.
- Decide per class: leaver accounts are disabled after owner confirmation; data errors are fixed at source; shared and admin accounts are assigned an accountable owner and moved under PAM; service accounts are recorded as non-human identities with an owner and purpose.
- Handle duplicates and ambiguity: one identity with several accounts in the same app is a review finding in itself, often a privileged second account.
- Re-run until clean: aim for zero unexplained accounts before the baseline review starts.
Non-human accounts found during onboarding deserve their own track rather than being forced into a human review. Record the owner, the system that uses the account, its credential type and rotation, and review it in an ownership attestation. See the NHI Ownership Guide. The Microsoft Midnight Blizzard breach shows the cost of the alternative: a legacy test application with elevated access to the corporate environment, outside normal governance, became the attacker's route to executive mailboxes.
Running the baseline review and going live
The baseline review confirms that access granted manually or out of band before onboarding is still needed. Microsoft's onboarding sequence puts it after integration and before deploying request policies, for the same reason: otherwise self-service starts from an over-provisioned state.
- Prepare the owner: confirm entitlement descriptions and risk ratings are in place and reviewers have been briefed.
- Run a full review of the application: every account and entitlement, reviewed by the manager for business need and by the entitlement owner for high-risk items, with the reviewer shown last-used dates where the app exposes them.
- Remediate and evidence: automatic removal where provisioning exists; tickets with closure evidence where it does not; confirm through the next import that revoked access is actually gone.
- Turn on policies: birthright mappings, request workflows with the catalogue's approval stages, expiry for time-bound access and SoD checks.
- Set the ongoing cadence: risk-based, for example quarterly for privileged entitlements and annually for low-risk ones, plus event-driven reviews on moves. See the Access Reviews Guide.
- Rollback: if a connector misbehaves in production, switch it to read-only aggregation and fall back to ticket fulfilment while it is fixed; never let a faulty connector keep deprovisioning.
Pilot new connectors against a non-production instance of the app, and run the first production aggregation read-only for at least one cycle before enabling writes. Treat the connector's own service account as a privileged non-human identity: vault its credential, scope it to the operations it needs and log its use.
Detection and measurement
Monitor for drift between the governed state and the application's real state: accounts created directly in the app, entitlements granted outside the request process, and revoked access that reappears. Each aggregation is a detective control; alert on new privileged entitlements that have no matching approved request.
- Coverage: percentage of risk-weighted applications onboarded, split by read-only and provisioning.
- Catalogue quality: percentage of entitlements with an owner, business description and risk rating (target 100% for high risk).
- Correlation rate: percentage of accounts matched to an identity or registered as an owned non-human account.
- Out-of-band grants: number of entitlements found by aggregation without an approved request, per month.
- Revocation closure time: median days from review revocation to confirmed removal, separately for connected and disconnected apps.
- Review quality: revocation rate and approve-all rate per reviewer, to spot rubber-stamping.
Framework mapping
| Framework | Reference | Onboarding activity |
|---|---|---|
| NIST SP 800-53 Rev 5 | AC-2 Account Management; AC-2(1) Automated System Account Management; AC-2(3) Disable Accounts | Correlation, orphan remediation, connector-driven lifecycle |
| NIST SP 800-53 Rev 5 | AC-5 Separation of Duties; AC-6(7) Review of User Privileges | SoD conflicts in the catalogue; baseline and periodic reviews |
| NIST SP 800-53 Rev 5 | CM-8 System Component Inventory | Application inventory that drives the backlog |
| NIST CSF 2.0 | ID.AM-02; PR.AA-01; PR.AA-05 | Inventory of systems; managed identities and credentials; entitlements defined, managed and reviewed |
| CIS Controls v8.1 | Safeguards 5.1, 6.1, 6.2, 6.8 | Account inventory, access granting and revoking processes, role-based access control |
| ISO/IEC 27001:2022 Annex A | 5.15, 5.16, 5.18, 8.2 | Access control, identity management, access rights, privileged access rights |
Common failure patterns
- Counting apps instead of risk. Reporting "120 apps onboarded" while the ERP, the payments platform and the cloud consoles remain outside governance.
- Importing raw entitlements unedited. Thousands of technical names flow into reviews, reviewers approve everything, and the audit evidence is worthless.
- Disconnected app revocations never verified. Tickets are closed but the next import is not checked, so revoked access survives indefinitely.
- Forcing service accounts into human reviews. Managers cannot judge them and approve or revoke blindly, breaking integrations or keeping unowned credentials alive.
- Turning on self-service before the baseline review. Automation then preserves and extends existing over-provisioning.
- Ignoring non-production and test instances. Test tenants and apps with links into production are left out of scope, as in the Microsoft Midnight Blizzard breach.
Practitioner checklist
- Build a risk-scored application inventory and plan onboarding in waves of similar integration type.
- Name an application owner and entitlement owners, with deputies, before any technical work.
- Run an entitlement workshop per app to write business descriptions and apply published risk criteria.
- Choose the lightest integration pattern now (read-only import is acceptable) and record the target pattern.
- Correlate on a stable identifier, classify every unmatched account, and reach zero unexplained accounts.
- Register service and shared accounts as owned non-human or privileged identities, not as human access.
- Complete a baseline review and confirm removals before enabling requests, birthright rules and SoD checks.
- Verify ticket-based revocations against the next import and track closure time.
- Vault and scope every connector service account and log its activity.
- Report coverage, catalogue quality, correlation and out-of-band grants monthly.
Standards and references
- NIST: SP 800-53 Rev 5 Security and Privacy Controls (AC-2, AC-5, AC-6)
- IDPro Body of Knowledge: Optimizing Access Recertifications
- IDPro Body of Knowledge: Techniques To Approach Least Privilege
- IETF: RFC 7644 System for Cross-domain Identity Management: Protocol
- Microsoft Learn: Govern access for applications in your environment
- Microsoft Learn: Define organisational policies for governing access to applications
- Microsoft Learn: Govern the users of an application that does not support provisioning
Related NHI Mgmt Group resources: IGA Buyer's Guide · Access Reviews Guide · Joiner-Mover-Leaver Guide · Segregation of Duties Guide · Entra ID Entitlement Management and Access Packages Guide