Decision rights / role clarity
RACI Matrix
RACI maps every decision or deliverable to four roles: who is Responsible for doing it, who is Accountable for the outcome (exactly one), who is Consulted before, and who is Informed after.
It is the fastest instrument for the two commonest org diseases — decisions with three owners, and decisions with none.
- Problem
- Decision rights / role clarity
- Altitude
- Team to enterprise
- Effort to run
- Light
- Evidence base
- Established
Theory & origin
Responsibility charting grew out of 1970s project-management practice (also known as a responsibility assignment matrix, with variants like RASCI and DACI). The rule that gives the tool its teeth is the single-A constraint: exactly one accountable name per row. Everything diagnostic follows from violations — multiple As mean a turf war in waiting, zero As mean an orphaned decision, a column drowning in Cs means a consultation culture where nothing moves. Critics rightly note a RACI is a snapshot that drifts as the organization changes; it documents agreements, it does not create the trust to keep them.
Explore the model
How a consultant runs it
- 01 Chart real decisions and deliverables, not departments — "approve pricing exceptions", not "pricing".
- 02 Enforce the single-A rule row by row; every argument it starts is a conflict the org was already having silently.
- 03 Read the columns as diagnosis: role overloaded with As, everyone consulted on everything, stakeholders informed of nothing.
- 04 Resolve the violations in the room with the people named — a RACI negotiated by proxy changes nothing.
- 05 Timestamp it and revisit on reorgs; a stale RACI is more dangerous than none because people still cite it.
When to use
- 01 After a reorg or merger, when nobody is sure who owns what anymore
- 02 Chronic decision gridlock: approvals with three signatures or deliverables with no owner
- 03 Standing up cross-functional processes where handoffs keep dropping
When not to use
- 01 As a substitute for org design — a RACI documents ownership, it cannot fix a structure that splits it
- 02 For fast-moving teams whose work changes weekly; the chart drifts faster than it can be maintained
- 03 Charted in the abstract by HR alone; unnegotiated letters change nothing on the ground
Worked example
A scale-up’s pricing exceptions take three weeks and everyone blames everyone. Charting eight commercial decisions exposes the pattern: "approve discount >15%" has three As (Sales VP, CFO, and legacy founder habit), and "update the price book" has none. The workshop assigns the CFO as the single A with a 48-hour SLA, sales VP moves to C, the price book gets an owner. Exception turnaround drops to two days, and the founder’s sign-off habit — the real blocker nobody would name — is retired explicitly rather than silently resented.
Common pitfalls
- 01 Multiple As per row, faithfully documenting the ambiguity the tool exists to kill
- 02 Charting departments instead of decisions, producing a wall poster nobody can act on
- 03 Consultation inflation: Cs handed out as diplomacy until every decision needs a summit
- 04 Filing the matrix and never revisiting it, so the org runs on a chart two reorgs old
Sample deliverable
One real engagement, end to end — watch the numbers travel from raw input, onto the chart, into the artifact.
Input — raw data
- Cutover date1 A ✓
- Data-migration signoff1 A ✓
- Branch training plan1 A ✓
- Rollback decision2 A’s ✗
Process — mapped
| CIO | CFO | Ops Dir | Vendor | |
|---|---|---|---|---|
| Cutover date | A | C | R | I |
| Data-migration signoff | C | A | R | R |
| Branch training plan | I | I | A | C |
| Rollback decision | A | A | R | I |
Each decision is charted role by role; double-A cells (clay border) surface the gridlock instantly
RACI — core-banking migration, BPR Nusantara
- Violationrollback has two A’s (CIO + CFO)
- FixCIO sole A, CFO moves to C, 48h SLA
- Logevery decision recorded against the chart