Glossary Term

Segregation of Duties (SoD)

The control principle that no single individual should be able to create, approve, and verify a GxP record without independent oversight — operationally enforced through RBAC.

If one person can create a record, approve its release, and verify the release, and nothing in the system blocks them, the record carries that doubt. Segregation of duties is the policy that closes the gap; RBAC is how the policy gets enforced.

Segregation of duties matrix showing role conflicts and required separation across GxP activities
On this page
  1. Definition
  2. Why It Matters
  3. Regulatory Context
  4. In Practice
  5. Key Controls
  6. Complere Approach
  7. Related Terms

What segregation of duties is

Segregation of Duties (SoD) is the control principle that critical activities affecting GxP records should be divided among different individuals so no single person can both perform the activity and approve it. The principle reduces fraud risk, error risk, and the bias of conflict-of-interest. In regulated electronic systems, SoD is the policy layer that role-based access control enforces operationally.

The classic regulated cases: an analyst running a test shouldn't approve their own result; a document author shouldn't approve their document for release; a change requester shouldn't approve the change; an investigator shouldn't approve closure of their own investigation; a system administrator shouldn't have routine GxP user privileges; an audit lead shouldn't both run the audit and close the findings.

SoD is sometimes called "separation of duties" or "the four-eyes principle." The terminology varies; the substance is the same. Regulators expect documented SoD analysis, role definitions that enforce it, compensating controls for unavoidable exceptions, and periodic review that catches drift.

SoD is the policy; RBAC is the enforcement

Without an SoD analysis, role design tends to create conflicts. Without RBAC, even a perfect SoD analysis can't be enforced sustainably. Programs need both: the SoD discipline that identifies which pairs of activities must be separated, and the role-based access infrastructure that prevents one user from holding both sides.

Why SoD has become an explicit inspection topic

SoD weaknesses turn up in nearly every data integrity inspection finding. The pattern: an analyst creates a test result, the same analyst electronically signs the approval, no one independent reviews the underlying data, and an OOS result later surfaces a data integrity issue the approval should have caught. The system permitted what the policy should have blocked.

Modern data integrity guidance from MHRA (March 2018, updated September 2021), PIC/S PI 041-1 (July 2021), and FDA's December 2018 DI Q&A all treat SoD as a baseline control. PIC/S PI 041-1 specifically references the four-eyes principle and the need for role separation. Inspectors look for SoD analysis documented in role definitions and access control SOPs, not just stated as policy intent.

Inspector perspective: a typical inspector technique is to pick one critical workflow — a batch release, a CAPA closure, a deviation disposition — and trace it back. Who initiated it? Who reviewed it? Who approved it? If the same name appears in two of those three positions, the inspector will want to see the SoD analysis that accepted that combination and the compensating controls that mitigate it. If the analysis doesn't exist, the conversation about whether the record can be trusted starts there.

Where SoD obligations come from

SoD isn't named in a single section of any predicate rule; the obligation comes from several intersecting expectations:

  • 21 CFR §11.10(d): limiting access to authorised individuals — presupposes that authorisation is granted by role, with conflicts considered.
  • 21 CFR §11.10(g): authority checks. The system checks the user has the authority for the specific action; the authority structure is what enforces SoD.
  • 21 CFR §11.100: unique signature attribution. SoD applies because a single user's signature shouldn't be acceptable for both sides of a separated activity.
  • 21 CFR §11.50: signature meaning. The meaning captured (review, approve, responsibility) is the layer SoD analysis operates on.
  • 21 CFR §211.22(a): quality control unit responsibility for approval or rejection — the predicate for QA independence from production.
  • 21 CFR §211.68(b): control over computerised systems including the people authorised to make changes.
  • EU GMP Annex 11 §2 — Personnel: clear definition of responsibilities.
  • EU GMP Annex 11 §12 — Security: authority to enter, change, or delete data should be defined; access should reflect role; expects SoD-aware role design.
  • EU GMP Chapter 2 §2.5: roles and responsibilities clearly defined to prevent overlap.
  • MHRA "GxP" Data Integrity Definitions and Guidance for Industry (March 2018, updated September 2021): addresses segregation of duties as a data integrity control.
  • PIC/S PI 041-1 (July 2021): explicitly references the four-eyes principle and role separation; sets the most detailed regulator expectation on the topic.
  • FDA Data Integrity and Compliance With Drug CGMP Q&A (December 2018): addresses SoD in the context of supervisor review and independent review of data.
  • ICH Q10 §2: management responsibility for resources and structure; SoD-aware role definition sits inside this.
  • ISO 13485 §5.5: responsibility, authority, communication — the device-side framework for documented role separation.
  • ISO 27001 A.6.1.2: segregation of duties as an information-security control; often referenced in regulated firms' broader access policy.

What a working SoD program contains

A defensible SoD program contains these elements operating together:

  • Documented SoD matrix. A controlled document identifying the activity pairs that must be separated, the roles that enforce the separation, and the rationale. Updated when roles change or new systems come online.
  • SoD analysis at role design. When a role is defined or revised, the SoD matrix is checked: does this role grant rights that conflict with another role? Can a single user hold both? If yes, what's the mitigation?
  • SoD enforcement at assignment. When a user is assigned to a role, the assignment workflow checks against other roles the user holds. Conflicting assignments trigger an approval gate.
  • Documented and approved exceptions. Where business need requires a conflict (small team, specialist function), the exception is documented with justification, compensating controls, senior management approval, and periodic re-confirmation.
  • Compensating controls actually performed. Independent secondary review, elevated audit trail review frequency, expanded periodic access review — defined for each exception and operationally executed.
  • Periodic SoD review. The SoD matrix and current role assignments reviewed periodically (typically annually) to catch drift. Role definitions that have expanded since the matrix was approved get flagged.
  • Periodic access review includes SoD check. When access is reviewed quarterly, the review explicitly checks for conflicting role combinations on individual users, not just whether each role is still warranted.
  • Audit trail review surfaces SoD violations. Patterns like the same user signing both review and approval of the same record are flagged for investigation.
  • Training on SoD principles. Users with multiple roles understand which combinations are constrained and why. Role owners understand their role's SoD implications.
  • Management Review input. SoD exception count, exception aging, compensating-control closure rate are standing inputs to Management Review.

What strong SoD programs do

The patterns that hold up at inspection:

The 'small team exception' trap

A common pattern: a small team can't fully segregate duties because there aren't enough people. The exception is taken, documented once, and forgotten. Years later, the team is larger but the exception remains in place. Periodic SoD review catches this — the exception is re-confirmed each year against current team size. Where the original justification no longer holds, the exception is closed and full segregation re-established.

  • SoD matrix is current. Versioned, approved, periodically reviewed; not a static spreadsheet from initial system go-live.
  • Role definitions designed against the matrix. Not retrofitted to the matrix after the fact.
  • System enforces, doesn't just suggest. The system blocks conflicting role assignment, not just warns. Override requires documented approval.
  • Exceptions are governed. Documented, justified, senior-management-approved, and time-bound or periodic-reviewed.
  • Compensating controls operate. Defined for each exception, performed on schedule, evidence retained.
  • Periodic SoD review separate from access review. Both happen; both produce records. The SoD matrix itself gets reviewed, not just the assignments against it.
  • Audit trail review surfaces SoD violations. Same-user-on-both-sides patterns get flagged in the review program.
  • Training reinforces SoD culture. Users with multiple roles know which combinations matter and why.
  • Management Review tracks SoD metrics. Exception count, exception aging, compensating-control effectiveness — all reviewed.
  • External and contractor access included. SoD constraints apply equally to vendors, consultants, and auditors.

How Complere supports SoD enforcement

SoD analysis is your work — Complere can't decide for you which activity pairs need separation in your firm's context. What the platform does give you is the role and authority infrastructure that turns your SoD policy from intent into enforcement, so the system blocks what your policy says shouldn't happen.

You configure roles per regulated workflow. Inside each workflow, roles map to specific actions — draft, review, approve, sign, close — and to the signature meanings each role is allowed to apply. That mapping is how you enforce, for example, that the person who reviewed a CAPA can't also approve it: you define "reviewer" and "approver" as distinct roles with distinct signing authority, and your assignment process keeps one user from holding both unless you've documented an exception.

When one of your users tries to act on a record, the system checks at that moment whether they have the action permission and the signature authority. Rights you revoked yesterday don't carry over today. Every signing event is recorded with the user, the meaning, the timestamp, and the binding to the record — so when your team or your auditor reviews the trail, any same-user-on-both-sides pattern shows up plainly.

What stays your work: the SoD analysis itself (which pairs of activities your firm separates, which conflict combinations need senior justification), the role design that operationalises that analysis, the documented exceptions and the compensating controls that govern them, your periodic SoD review cadence, and the training that keeps your team aware of which combinations matter and why. Complere supports the program; you run it.

Frequently asked questions

Common questions about Segregation of Duties (SoD) sourced from regulatory references and inspection patterns.

What is segregation of duties in regulated quality?

SoD is the control principle that critical activities affecting GxP records should be divided among different people so no single individual can both perform the activity and approve it. The classic case: an analyst who runs a test shouldn't also approve their own result. A document author shouldn't also approve their document for release. A change requester shouldn't also approve the change. The principle reduces fraud risk, error risk, and the bias of conflict-of-interest.

What activities specifically require SoD?

The common pairs that need separation: data creation and data approval; change initiation and change approval; deviation investigation and deviation closure; audit execution and audit response; system administration and routine system use; supplier approval and supplier audit; record creation and record review. Programs typically maintain a documented SoD matrix listing conflict pairs and the role-level separations that enforce them.

What's the relationship between SoD and RBAC?

SoD is the policy; RBAC is the enforcement mechanism. The SoD analysis identifies which activity pairs need separation. The role definitions in RBAC then ensure no single role holds both sides of a conflict pair, and the assignment rules ensure no single user holds two conflicting roles. SoD without RBAC tends to rely on procedural controls (training, signed acknowledgements) that fail under inspection.

What if full SoD isn't operationally feasible?

Small teams, specialist roles, or low-volume regulated activities sometimes can't achieve full SoD. The accepted approach: identify the conflict, document why segregation isn't feasible, implement compensating controls (independent secondary review, increased audit trail review frequency, expanded periodic access review), and have senior management sign off. Uncontrolled SoD exceptions are non-compliant; controlled-and-documented exceptions are typically accepted.

What are compensating controls?

Controls that mitigate the residual risk when full segregation isn't possible. Examples: an independent quality reviewer signs off on records the conflicted user creates and approves; audit trail review on conflicted-role actions runs at higher frequency (weekly instead of monthly); periodic access review on conflicted-role assignment runs more often; the conflict is brought to Management Review for awareness. The point is that the residual risk is visible, governed, and accepted.

Where does SoD obligation come from in Part 11?

Not from a single section. 21 CFR Part 11 §11.10(d) restricts access to authorised individuals; §11.10(g) requires authority checks; §11.100 requires unique attribution; §11.50 requires signature meaning. Together these expect that the user signing has the authority to apply that meaning, which presupposes SoD analysis at the role-definition level. EU GMP Annex 11 §12 is more explicit, defining access by role with authority differentiation.

What are the most common SoD findings in inspections?

Shared accounts (which defeat SoD by hiding which user took which action); a single user holding conflicting roles; SoD analysis missing or out of date; SoD exceptions undocumented or unjustified; compensating controls listed in policy but not actually performed; periodic access review missing the SoD check entirely. The chromatography enforcement actions that ran from 2017 onward drove several of these patterns.

How often should the SoD analysis be revisited?

Whenever roles change, when new systems come online, when new activities are added to existing systems, when organisational structure shifts (mergers, divestitures, team reorganisations), and on a periodic cadence (typically annual) even when nothing has obviously changed. Programs that only do SoD at initial system implementation tend to find drift on first inspection.

About the author

Complere Reference Team

Compliance and quality-systems specialists maintaining the Complere glossary for regulated quality, validation, and inspection-readiness teams. Entries are reviewed against current FDA, MHRA, EMA, ICH, and PIC/S guidance.

Continue Exploring

Explore related topics, modules, and compliance resources for a deeper understanding of your quality system.

Data integrity hub
Related

Data Integrity & Audit Trails

Explore this topic in more depth to build a complete picture of your quality and compliance operations.

Explore
Electronic records and signatures
Related

Electronic Records & Signatures

Explore this topic in more depth to build a complete picture of your quality and compliance operations.

Explore
Data Integrity glossary
Related

Data Integrity

Explore this topic in more depth to build a complete picture of your quality and compliance operations.

Explore

See segregation of duties enforced in Complere

Walk through Complere's role-based access, role-to-signature mapping, and workflow stage controls that operationally enforce the SoD policy.