Skip to content

RACI or Decision Participation Matrix

Coordination tool — instantiates Role Expectation Architecture

Lays every recurring task or decision against every role in a grid and tags each cell, so exactly one role is Accountable and no decision right is left blank or doubled.

RACI (Responsible, Accountable, Consulted, Informed) is a grid: recurring decisions down one side, roles across the top, a tag in every cell. Its distinctive move is working across many roles at once against a shared list of decisions, under one iron rule — exactly one Accountable per row. That rule is what turns "who owns this?" from a standing argument into a lookup. The matrix says nothing about what any role is or does day to day; it answers only the narrow, coordination-critical question of who may decide each item and who answers for it. Read a row to see a decision's participants; read a column to see one role's total decision load.

Example

A mid-rise construction project keeps stalling on change orders. The architect assumes the general contractor will price the change, the GC assumes the owner's representative will approve the scope, and the structural engineer only hears about load-bearing changes after they are built. Rather than redesign anyone's job, the project manager builds a RACI over the dozen recurring decisions. For approve a structural change, the row resolves to: structural engineer = R (does the analysis), owner's rep = A (one signature; the decision stops here), architect and GC = C (consulted before), the trades = I (told after). Filling the grid exposes two pathologies at once: three rows carry two Accountables — that is the standoff — and one row, notify the city inspector, carries none — the gap nobody owned. The remedy is not new roles; it is assigning each row a single A. The change-order backlog clears because every decision now has exactly one desk it stops at.

How it works

The unit of the tool is the decision, not the role. You enumerate the recurring decisions first, then tag each role's participation per decision: R does the work, A owns the outcome (one and only one per row), C must be consulted before, I is informed after. The whole method rides on two disciplines — the single-A rule and the no-blank-row rule — because a row with two Accountables is a turf collision and a row with none is an orphaned decision. The grid is a reconciliation across roles: its value comes from being read both ways, down columns to catch overload and across rows to catch gaps.

Tuning parameters

  • Task-list granularity — a handful of high-stakes decisions versus every routine action; finer rows catch more gaps but bloat the grid into unread shelfware.
  • Single-A strictness — whether you genuinely permit only one Accountable per row (recommended) or tolerate shared accountability; loosening it re-admits the very ambiguity the tool exists to remove.
  • C-versus-I threshold — how much involvement counts as consulted (input or veto before) rather than merely informed (told after); set it too generously and every decision drowns in mandatory consultation.
  • Variant scheme — plain RACI versus extensions (RASCI adds Support; RACI-VS adds Verifier and Signatory); add letters only when a real distinction is otherwise being lost.
  • Refresh trigger — one-time snapshot versus a re-run whenever the org or the workflow shifts; a stale matrix quietly misroutes decisions long after it stops matching reality.

When it helps, and when it misleads

Its strength is making coordination gaps and overlaps visible on a single page and converting "who owns this?" into a lookup; the single-A rule is a direct antidote to decisions that stall because everyone — or no one — owns them. Its failure modes are real: it can calcify into bureaucracy, a cell for every trivial act, and it flattens time — it says who participates, never in what order, so a "clean" RACI can still hide a broken handoff sequence. Its classic misuse is being filled in after a turf outcome to ratify who already grabbed what, rather than to design participation deliberately. The pathology it exists to guard against is diffusion of responsibility — the more people nominally share a duty, the less any single person acts on it[1] — and the one-Accountable-per-row rule is precisely the structural countermeasure. The discipline that keeps it honest: keep the row list lean, enforce a single A, and re-derive the grid from the actual workflow rather than from the current power map.

How it implements the components

RACI fills the decision-rights slice of the architecture — who may decide and who answers — and nothing about what the role is:

  • authority_and_permission_scope — each row pins the decision right: the A (and often the R) is where the authority to decide that item actually sits, made explicit and non-overlapping.
  • accountability_and_review_path — the single Accountable per decision is the accountability anchor and the first stop for review or escalation when that decision goes wrong.

It does not model what a role does day to day (expected_behavior_bundle) or its identity (role_position_anchor) — that is the Role Card; it does not map the standing service flow and handoffs between roles (role_interface_map, obligation_and_service_scope) — that is the Swimlane or Service Blueprint; and it is not the instrument that formally grants an individual their authority — that is the Delegation Letter or Authority Envelope.

  • Instantiates: Role Expectation Architecture — RACI supplies the decision-rights-and-accountability layer of the role design.
  • Consumes: Position Description or Office Mandate — supplies the defined roles the matrix arranges into decision participants.
  • Sibling mechanisms: Swimlane or Service Blueprint · Role Card or Participation Card · Role Charter · Delegation Letter or Authority Envelope · Position Description or Office Mandate · Role Review Retrospective · Role Rotation or Deputy Schedule · Handoff Checklist · Onboarding and Role Shadowing Runbook · Conflict-of-Interest Disclosure · Role Compatibility Check

References

[1] Diffusion of responsibility — the finding from social psychology (Darley and Latané's bystander research) that individuals feel and act on less personal responsibility as the number of others nominally responsible grows. A matrix that allows several Accountables per decision re-creates exactly this effect; the one-A-per-row rule is the standard structural guard against it. ↩