Home›Blog›RACI matrix: the four roles, and where the grid comes from
FundamentalsBeginner
Published September 6, 2026 · By Vincent KENNEL
RACI matrix: the four roles, and where the grid comes from
Four letters, and the referentials do not gloss them the same way. What the acronym actually settles is narrower than it looks, and where the grid came from is a shorter story than the one that circulates.
In brief
A RACI matrix is a responsibility assignment matrix in which each cell carries a letter naming the role held on one task: R does the work, A answers for the result and there is only one, C is consulted, I is informed. The PMBOK Guide, 3rd edition (2004), records the format.
Put several organisations on the same piece of work and a question appears that no organisation chart answers: when the thing is not done, who answers for it? An organisation chart says who reports to whom. It says nothing about who carries a given task, whose opinion has to be collected before it moves, and who only needs to hear about it afterwards. A RACI matrix exists to close that gap, one task at a time.
What a RACI matrix is
A RACI matrix is not a tool of its own. It is a way of filling in a matrix that exists without it. The PMBOK Guide, 3rd edition (2004), where the acronym enters the standards, names both objects together in the title of its figure 9-5: a responsibility assignment matrix using a RACI format. What the acronym adds is not the table, which was already there. It is the content of the cell.
Three other traditions say the same in their own vocabulary. The PMBOK Guide, 4th edition (2008), is the bluntest: RACI is one type of responsibility assignment matrix among others. AACE International Recommended Practice 83R-13 (rev. 1 May 2014) has a matrix that may be further developed into a RACI chart. The APM Body of Knowledge, 8th edition (2025), treats RACI as a coding applied to a matrix defined elsewhere, never as an instrument. The shape is stable throughout: work items in rows, roles in columns, one letter per cell. The PMI Guide to Business Analysis (1st edition, 2017) publishes that shape as a reference format.
Nobody imposes it either. PRINCE2 7 (AXELOS, 2023) files RACI among supporting techniques, outside the method itself; the AACE practice calls it an optional refinement.
One boundary before going further. In the earned value tradition the controlling matrix carries a cross and then amounts, not letters, and the AACE practice is explicit that only the R designation corresponds to the control account manager responsible for the work. A RACI and a responsibility assignment matrix do not point at the same person. The matrix itself, and what crossing work with organisation reveals, belong to a separate article.
The four letters, referential by referential
PRINCE2 7 gives the only gloss in this corpus that states cardinality outright. Responsible: one or more people who perform the task. Accountable: the single person who owns it. Consulted: those whose input is required. Informed: those told of progress or completion.
Only one source explains the split rather than stating it. COBIT 4.1 (IT Governance Institute, 2007) glosses accountable as the point where the buck stops, the person who gives direction and authorises the activity, and responsibility as belonging to whoever gets the task done.
The line between C and I is the best evidenced point of the four, and it is not a degree of involvement. It is the direction of the communication. The BABOK Guide v3.0 (IIBA, 2015) has the communication running one way with informed and both ways with consulted. The PM² Project Management Methodology Guide 3.1 (2023) reaches the same criterion independently. So do Michael L. Smith and James Erwin in Role and Responsibility Charting (RACI), where consult is a predetermined need for two-way communication and inform is one-way. Three traditions with no shared lineage, one criterion.
And an absence. Across seven volumes, from the PMBOK Guide, 1996 edition, to the PMBOK Guide, 7th edition (2021), PMI expands the acronym and never glosses the four letters one by one. Its only definition is of RACI as a block, which entered the glossary with the PMBOK Guide, 5th edition (2013). A reader looking for the PMI definition of the four roles is looking for something that does not exist.
Nor is the dispersion confined to PMI. The BABOK Guide v3.0 carries three versions in a single volume, and the APM Body of Knowledge, 8th edition (2025), two a page apart. That is why every definition here names the edition it comes from. On this subject, naming the edition is not a citation habit. It is the finding.
R and A, the only thing the tool decides
Three letters out of four describe participation. One creates an obligation.
That one is A, and it is the single point on which the referentials do not diverge. PRINCE2 7 has the one person who owns the task, the BABOK Guide v3.0 the one stakeholder who receives the assignment, PM² just one accountable person per task. Rita Mulcahy, in PMP Exam Prep (9th edition, 2018), holds only one person accountable, and Smith and Erwin allow only one A per action.
On R there is no such agreement. PRINCE2 7 admits one or more, Mulcahy allows several, PM² requires exactly one, and Smith and Erwin have responsibility that can be shared, with its degree set by whoever holds the A. There is no universal rule to apply here. There is a referential being followed, and it has to be named.
Figure 1: two partitions, not four roles in a queue
In everyday English, responsible and accountable are used almost interchangeably, and the matrix exists precisely to prise them apart. Responsible is the one who does the work. Accountable is the one who answers for it when it is not done. A reader who treats the two as synonyms can fill in the whole grid and change nothing.
What the matrix makes visible
An example, deliberately generic: a team rolls out a new tool. Six work items, four roles.
Work item
Project manager
Business lead
Technical team
Executive management
Gather the requirements
A
R
C
I
Choose the solution
R
C
C
A
Configure the tool
A
C
R
I
Run acceptance testing
C
A
R
I
Train the users
A
R
C
I
Go live
R
C
C
A
A row reads the way everyone reads it. For one piece of work: who does it, who answers for it, whose input is needed, who gets told afterwards.
A column is the reading almost nobody does, and it costs nothing, since the grid is already filled. Read top to bottom, it says what one role carries across the whole of the work. Executive management appears twice as A and never as R; the technical team, twice as R and never as A. Neither fact is visible in a list of roles.
The line that teaches most is the A itself. It sits with executive management on choosing the solution and with the project manager on configuring. It has moved between two rows of the same grid, which means accountability here is not a rank in the hierarchy. It is attached to a task, and it changes with the task. A list of job titles cannot show that. A matrix shows it at a glance.
One closing remark, not developed here. When the R sits with a supplier and the A with the client organisation, the grid crosses a contractual boundary and the letters stop being an internal convention.
RACI, RASCI, and the rest
Variants circulate by the dozen. One has a public normative basis.
RASCI is the canonical form of PM², the project management methodology of the European Commission, which presents the responsibility assignment matrix as also known as a RASCI table and never uses the four-letter acronym in any of the five guides of that family held here. Support is defined as a role that works with the person responsible and helps complete the task, and the line with consulted is drawn from the other side: the consulted role does not help complete the task.
What that buys is a reading of three definitions rather than a sentence any of them writes. PM² requires exactly one R and defines S as the role working alongside that R. The S is what lets the R stay singular, where referentials without an S have to admit several. RASCI is not RACI plus a letter. It is a different distribution.
One further name is on record: PMI, in Requirements Management: A Practice Guide (2015), notes that the approach is known as either RACI or ACRI. Beyond those, a full sweep of this corpus finds no variant with any normative basis at all. The forms circulating online rest on books, or on nothing.
Where RACI comes from
Start with the shape. In September 1954, Dun's Review and Modern Industry published Alfred G. Larke on the linear responsibility chart, a tool introduced by the Serge A. Birn Company of Louisville, Kentucky and recorded by two deposits in the Catalog of Copyright Entries of the Library of Congress, dated 1 December 1953. The method has not changed since: positions along one axis, functions along the other, a symbol at the intersection. And already in that article, one company had replaced the deposited symbols with its own. The notation was unstable from the first day it was described.
The lineage usually cited starts later, with David I. Cleland and Wallace Munsey in the Harvard Business Review in 1967. Larke precedes it by thirteen years.
The documented grid closest to RACI is elsewhere again. Richard Beckhard and Reuben T. Harris set out responsibility charting in Organizational Transitions: Managing Complex Change (1st edition, 1977), and again in the 2nd edition (1987). Four codes: R for responsibility, A-V for approval required or the right of veto, S for support, I for inform. Three things there are not what a reader of RACI expects. The rule of one bears on the R, not on the A. The A is a veto, not an accountability. And there is no C at all, the I of the 2nd edition (1987) covering those who must be informed or consulted before action but cannot veto. RACI does not swap an S for a C. It splits one letter in two and drops support.
Neither edition claims the invention. The 1st edition (1977) introduces the technique as one that has emerged in recent years, the 2nd edition (1987) as one that has been developed, and there is no note, no bibliography and no attribution in either volume.
Figure 2: four correspondences between two documented grids
Then the acronym itself. In 1988, in a peer-reviewed management journal, George Korey still called the technique a linear responsibility chart, and a sweep of indexed scholarly literature returns no work on RACI as a responsibility matrix before 2002. The PMBOK Guide, 1996 edition, and the PMBOK Guide, 2000 edition, contain no occurrence of RACI at all, and their responsibility matrix figure uses a different set of codes: P, A, R, I and S.
The first attestation in this corpus is the PMBOK Guide, 3rd edition (2004), page 206, in three occurrences that all sit on that one page, with no glossary entry, no acronym list entry and no index entry. A referential that adopts an acronym without explaining it is recording a usage it assumes its reader already knows. That is an indication rather than proof of anteriority, and it is offered as one.
Date
Piece
Does the acronym appear?
December 1953, September 1954
Serge A. Birn copyright deposits; Larke, Dun's Review and Modern Industry
No. Linear responsibility chart
1967
Cleland and Munsey, Harvard Business Review
Article not consulted. The literature cites it for the linear responsibility chart
1977 and 1987
Beckhard and Harris, Organizational Transitions, both editions
No. R, A-V, S, I
1988
Korey, Management Decision
No. Still linear responsibility chart
1996 and 2000
PMBOK Guide, 1996 and 2000 editions
No. Codes P, A, R, I, S
Up to 2002
Indexed scholarly literature
No
2004
PMBOK Guide, 3rd edition, figure 9-5, page 206
Yes. First attestation in this corpus
Between 1988 and 2004 nothing in this corpus fills the gap. That is the state of the evidence, and it is worth saying plainly rather than rounding it to a decade.
What circulates instead does not survive a check. One origin sentence appears word for word across several sites, all of them pointing at an article that does not contain it, and four further accounts coexist and contradict one another without a source between them. The most widely distributed RACI document of all, the fourteen pages by Smith and Erwin, has no historical section, no bibliography and no reference of any kind. It does not even carry a date.
In short
Two things are worth carrying away, and neither is a summary.
Before filling a cell, ask who answers if the work is not done. That is the only question the tool settles. Everything else it records, it records as participation, and participation was never what was in dispute.
And when reusing a definition of the four letters, name the edition it came from. The definitions do not agree, and the disagreement is not noise from the market. It is in the standards themselves, sometimes twice in the same volume.
Stop guessing. See the real impact.
Frequently asked questions
Q.RACI, RASCI, RACI-VS: does the variant matter?
A grid is only worth the agreement around it: what matters is that everyone reads it the same way, not how many letters it carries. Only RASCI rests on a published text whose roles can be argued over, the PM² guide 3.1 (2023). RACI-VS and the others are house conventions.
Q.Does every project need a RACI?
The useful question is not whether the tool is required, but when the responsibility is worth writing down. A grid earns its place as soon as several entities contribute to the same work and no single chain of command covers them all. No referential in this corpus requires one.
Q.Does a RACI replace a responsibility assignment matrix?
No. A responsibility assignment matrix exists without RACI and can be filled with other codes entirely. RACI is one filling convention among several, and the matrix itself, built by crossing work with organisation, is a separate subject.
References
AACE International - AACE International Recommended Practice 83R-13 - Organizational Breakdown Structure and Responsibility Assignment Matrix - Rev. 1 May 2014
