Most identity governance projects that fail do so at the HR integration, not in the governance tool: the events coming from HR are late, ambiguous or wrong, and everything downstream inherits the error. A leaver who is never flagged keeps access, a rehire gets a second account, a contractor converting to employee loses everything on day one. This guide is about designing the integration itself: which system is authoritative for which attribute, how lifecycle events are detected, how future-dated hires, rehires and conversions are handled, and how to test and monitor the flow. It is written for IAM engineers, IGA architects and HRIS analysts. For the joiner, mover and leaver process around it, see the Joiner-Mover-Leaver Guide.
Key takeaways
- Authority is per attribute, not per system: HR owns employment status, dates and manager; the identity platform owns username and email; write back what HR needs rather than letting two systems both edit a field.
- The immutable identifier decides everything: matching on a stable worker ID that survives rehire and conversion avoids duplicate accounts and orphaned ones; names and email addresses are never matching keys.
- Effective dates are where the bugs live: future-dated hires, back-dated terminations and future-dated conversions each need an explicit rule, or the connector will act too early or too late.
- Out of scope is not the same as terminated: a scoping filter change can disable thousands of accounts; enable protections such as skipping out-of-scope deletions and test every scope change.
- The integration account is privileged: the HR integration user and the provisioning credentials can create and enable accounts, so treat them as high-value non-human identities.
Authoritative sources and attribute ownership
"HR is the source of truth" is too vague to build from. Most organisations have several sources: a core HR system for employees, a vendor management system for contractors, a student information system in education, and sometimes a separate HR instance for an acquired company. Each source should be authoritative for a defined population and a defined set of attributes. Write this down as an attribute ownership matrix before configuring anything.
| Attribute | Authoritative source | Flows to | Notes |
|---|---|---|---|
| Worker ID (immutable) | HR or vendor management system | Directory employeeID, IGA identity key | Matching attribute; never reused for a different person |
| Employment status, hire date, termination date | HR | Directory enabled state, IGA lifecycle state | Drives joiner and leaver events; keep the effective date, not just the status |
| Legal and preferred name | HR | Display name, given name, surname | Preferred name for display; legal name only where required |
| Manager | HR | Directory manager, IGA approver | Manager changes trigger reviews and approval routing |
| Department, cost centre, location, job code | HR | Directory attributes, role rules | Inputs to birthright access; validate values against a reference list |
| Worker type (employee, contractor, intern) | HR or vendor management system | IGA identity type, scoping | Determines policy, end-date requirement and sponsor |
| Username, UPN, primary email | Identity platform | Written back to HR | Generated once, uniqueness enforced by the identity platform |
| Work phone | Telephony or identity platform | Optionally written back to HR | Pick one owner |
The SCIM enterprise user extension (RFC 7643) gives a common vocabulary for several of these, including employeeNumber, costCenter, organization, division, department and manager, which helps when the same data must reach SaaS applications. See the SCIM Provisioning Guide.
Identifiers and matching
The single most consequential design decision is the matching key: the attribute the provisioning engine uses to decide whether an HR record corresponds to an existing account. Microsoft's cloud HR planning guidance uses the employee ID by default (Workday WorkerID mapped to the Active Directory employeeID attribute), supports several matching attributes with a precedence order where evaluation stops at the first match, and recommends indexing employeeId in Active Directory before the first full synchronisation.
- Use a person-level identifier, not a record-level one: some HR systems hold several records for one person (one per employment or contingent engagement). Workday, for example, distinguishes the WorkerID from the WID that identifies an individual worker profile; profiles that share a WorkerID can then be treated as one identity.
- Never match on name or email: names change and collide; email is generated by the identity platform and is therefore downstream of the match.
- Decide the reuse policy with HR: if HR issues a new ID on rehire, the connector will create a new account; if it reuses the old one, it will re-enable the old account. Microsoft documents both behaviours. Choose one deliberately.
- Handle pre-existing accounts: before go-live, back-fill the employee ID onto existing directory accounts and resolve duplicates, or the first full sync will create a second account for many staff.
Event types and how they are detected
HR systems record business processes; identity platforms need events. The integration translates one into the other, either by polling for changes or by receiving pushed events.
| Lifecycle event | HR signal | Identity action | Design question |
|---|---|---|---|
| Pre-hire | Hire business process completed with a future effective date | Create account disabled, or enabled with a start-date gate | How many days ahead, and is the account usable before day one? |
| Joiner (start date) | Status becomes active | Enable account, assign birthright access, issue first credential | Which time zone defines the start of the day? |
| Mover | Job, department, location or manager change | Recalculate roles, schedule removal of old access, trigger review | Which attribute changes count as a move? |
| Leave of absence | Leave status with return date | Suspend sign-in or keep with restrictions | Does policy differ for long-term leave? |
| Leaver | Termination with effective date and time | Disable, revoke sessions and tokens, schedule deletion | Immediate termination path for dismissals? |
| Rehire | Hire for a worker with prior history | Re-enable old account or create new one; reset access to birthright | Old entitlements must not come back |
| Conversion | Contingent worker becomes employee, or the reverse | Transfer account to new profile or create new persona | Same account or new one, and on which date? |
| Rescinded offer | Hire cancelled before start | Delete or disable the pre-hire account | Is cancellation a distinct event in HR? |
Polling: the Microsoft Entra Workday connector, for example, runs an initial full sync that pages through hire and contingent worker business processes, saves a watermark timestamp, then runs incremental cycles that query manual updates, effective-dated changes and terminations within the window since the last run, plus specific lookups for future-dated hires. Push: Microsoft's API-driven inbound provisioning accepts SCIM-formatted bulk uploads at a /bulkUpload endpoint from any system of record, returns 202 Accepted and processes asynchronously. Microsoft recommends sending complete records each time and up to 50 operations per call, and documents limits of 40 calls in any five seconds and a daily tenant quota (2,000 calls with Entra ID P1 or P2, 6,000 with Entra ID Governance). Push gives faster leavers; polling is simpler to operate. Either way, the event handler must be idempotent: processing the same record twice must not create two accounts or two access requests.
Pre-hire, rehire and conversion patterns
Pre-hire
Creating accounts before the start date lets equipment be shipped and mailboxes exist on day one. The risk is a usable account for someone who has not started, or whose offer is withdrawn. In Workday, future hires show as inactive until the hire date; Microsoft's connector queries them as of the hire date so the profile can be pre-populated, and suggests a DateDiff expression where the account should stay disabled until the start date. A sound pattern is: create disabled N days ahead (commonly 7 to 14), enable at the start of the start date in the worker's time zone, and issue the first credential only after identity verification on day one. Fraudulent hires are a real risk here; see the Insider Threat from Fake IT Workers Guide and the Identity Proofing and KYC Guide.
Rehire
The two failure modes are a duplicate account (new ID, old account still present) and resurrected access (old account re-enabled with old group memberships). Strip entitlements to birthright at termination, or at re-enable, so a rehire starts clean, and run a manager-approved access request for anything beyond that.
Conversion
Contractor to employee conversions are the event most often handled by hand. Microsoft documents several Workday scenarios: back-dated and same-day conversions are picked up in the next cycle without changes, a worker who leaves and returns after a gap can have account ownership transferred to the new profile, but a future-dated conversion for an active contractor is processed immediately, updating the account with employee details early. The documented workaround is a Workday provisioning group for future-dated conversions, excluded with a scoping filter until the effective date. Microsoft's planning guidance also describes the alternative persona model, where a new account is created and the old one disabled on the conversion date. See the Contractor Identity Lifecycle Guide.
Scoping, error handling and safety rails
- Scoping filters: define which HR records are in scope (for example employees and contractors, excluding historical records). Microsoft warns that users falling out of scope are disabled or deleted by default, and provides a "skip out of scope deletions" setting to prevent accidental mass deactivation.
- Restrict target actions: limit the connector to create and update where deletion should be a separate, reviewed step.
- Quarantine and thresholds: Microsoft places a job in quarantine on a high error rate and disables it after more than four weeks in quarantine. Add your own threshold: stop and alert if a single cycle would disable more than a set number or percentage of accounts.
- Re-sync triggers: changing attribute mappings or scoping filters triggers a new initial cycle. For tenants above 30,000 users Microsoft recommends progressive rollout by widening the scope after each validation.
- Data quality checks: reject records with missing manager, invalid cost centre or a termination date before the hire date, and route them to an HR data steward queue rather than provisioning a half-formed identity.
- Personal data: provisioning logs and agent event logs contain HR data. Microsoft notes the provisioning service keeps data for 30 days and that agent Windows event logs may contain personal data; set retention accordingly.
Where identity fits: the integration's own credentials
An HR-driven integration is a set of non-human identities with power to create, enable and modify every workforce account. In Workday, the connector uses an Integration System User (ISU) that Microsoft's tutorial configures with "Do Not Allow UI Sessions", placed in an integration system security group with Get permissions on worker data domains and Put permissions only where writeback is used. On the directory side, the provisioning agent uses a group managed service account. For push models, the API client holds the SynchronizationData-User.Upload permission. Protect each of these as a privileged credential: least privilege on HR domains, constrained scope where available, vaulted secrets, monitoring for use outside the integration path, and an owner. Abuse maps to MITRE ATT&CK T1098 Account Manipulation and T1078 Valid Accounts. See the Group Managed Service Accounts (gMSA) Guide.
Implementation approach
- Agree ownership: HR, IAM and the service desk sign the attribute ownership matrix and the event definitions, including time zone and effective-time rules.
- Clean the data: back-fill worker IDs to existing accounts, resolve duplicates, decide what to do with accounts that have no HR match (service accounts, shared mailboxes, test users) and move them out of scope.
- Build in a test tenant or test app: Microsoft suggests separate test and production provisioning apps (for example "HR2AD (Test)" and "HR2AD (Prod)") so mapping changes are validated first; mask personal data in test copies.
- Run the scenario suite: new hire, future-dated hire, termination, immediate termination, manager change, department move, rehire with same ID, rehire with new ID, conversion in both directions, rescinded offer, name change.
- Pilot with a narrow scope: one department or location, read-only or create-only first, then enable updates, then disablement.
- Widen scope in steps and compare the provisioning log against the expected changes at each step.
- Rollback plan: stop the job, restore previous mappings from an exported configuration, and re-enable any accounts disabled in error from the provisioning log. Practise it once before go-live.
Decision guidance
| Choice | Option A | Option B | Choose A when |
|---|---|---|---|
| Integration style | Native connector polling HR | Push via API or middleware | A supported connector exists and leaver latency of a cycle is acceptable |
| Rehire identity | Reuse old account | New account | Audit continuity matters and access is reset at termination |
| Conversion | Transfer one account | New persona account | Users need continuity of mailbox and files and policies do not depend on persona |
| Pre-hire state | Disabled until start | Enabled with restrictions | Almost always; enable early only for a documented reason |
| Contractor source | Same HR system | Separate vendor management feed | HR already holds contingent workers with sponsors and end dates |
Small organisations with one HR system and a cloud directory can usually rely on the HR vendor's or identity provider's native connector and a simple matrix. Large organisations with several HR instances, contractors in a separate system and multiple directories need an identity hub (an IGA platform or a staging database) that merges sources by the person-level identifier before anything reaches target systems.
Detection and measurement
Monitor the provisioning logs, not just the HR system. Alert on: a cycle with more disables than the threshold; the job entering quarantine; sign-ins by accounts in a pre-hire or terminated state; accounts created in the directory without a worker ID; and changes to the connector's mappings, scope or credentials.
- Leaver latency: time from termination effective time to account disabled, reported as median and 95th percentile.
- Orphan rate: enabled human accounts with no matching active HR record.
- Duplicate rate: people with more than one active account.
- Day-one readiness: share of joiners able to sign in at their start time.
- Data quality backlog: HR records rejected by validation and their age.
- Manual intervention rate: lifecycle events needing a service desk ticket to complete.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev. 5 | AC-2, AC-2(3), PS-4, PS-5, PS-7 | Account management, disabling accounts, termination and transfer, external personnel |
| NIST CSF 2.0 | PR.AA-01, PR.AA-05, GV.RR-04 | Identities managed, access defined and reviewed, cybersecurity in HR practices |
| CIS Controls v8.1 | Control 5 (Account Management), Control 6 (Access Control Management) | Account inventory, disabling dormant accounts, access granting and revoking processes |
| ISO/IEC 27001:2022 Annex A | 5.16, 5.18, 6.5 | Identity management, access rights, responsibilities after termination or change |
| MITRE ATT&CK | T1078, T1098 | Use of accounts that should be disabled; manipulation of accounts via the integration |
Common failure patterns
- Termination recorded late: HR enters the termination after the last day, so the leaver keeps access for days. Fix the process with HR and add a manager-initiated urgent leaver path. See the Employee Offboarding Security Guide.
- Scope change disables thousands: a filter edit drops a population out of scope and the connector disables them in one cycle, because the default treats out of scope as deprovision.
- Future-dated change applied today: conversions or transfers entered in advance flow immediately, giving new access early or removing old access too soon.
- Two masters for one field: the service desk edits department in the directory and HR overwrites it nightly, so role rules flap and access is granted and removed repeatedly.
- Rehire resurrects old access: the re-enabled account still holds groups from the previous employment.
- Integration credentials over-scoped: the HR integration user has broad rights or interactive sign-in, making it an attractive target.
Practitioner checklist
- Publish an attribute ownership matrix signed by HR and IAM.
- Match on an immutable, person-level worker ID and back-fill it to existing accounts.
- Define every lifecycle event with its HR signal, effective-time rule and identity action, including rescinded offers and conversions.
- Create pre-hire accounts disabled and enable them on the start date in the worker's time zone.
- Reset access to birthright on termination so rehires start clean.
- Enable out-of-scope deletion protection and a mass-change threshold that stops the job.
- Run the full scenario suite in a test app before every mapping or scope change.
- Restrict, vault and monitor the HR integration user, provisioning agent account and API client.
- Report leaver latency, orphan rate and duplicate rate monthly.
Standards and references
- Microsoft Learn: Plan cloud HR application to Microsoft Entra user provisioning
- Microsoft Learn: How Microsoft Entra provisioning integrates with Workday
- Microsoft Learn: Configure Workday for automatic user provisioning
- Microsoft Learn: API-driven inbound provisioning concepts
- IETF: RFC 7643 System for Cross-domain Identity Management: Core Schema
- NIST: SP 800-53 Rev. 5
- NIST: Cybersecurity Framework 2.0
Related NHI Mgmt Group resources: Joiner-Mover-Leaver Guide · SCIM Provisioning Guide · Contractor Identity Lifecycle Guide · Entitlement Catalogue and Application Onboarding Guide · Identity Security Programme Guide