Org, people & execution
RACI Matrix
A grid that assigns one of four roles to everyone involved in a task or decision: Responsible for doing the work, Accountable for the outcome, Consulted before it is done, Informed after. Its whole discipline lives in one rule, exactly one Accountable name per row.
Also known as Responsibility assignment matrix, RAM, Responsibility charting, Linear responsibility chart. First set out by No single originator; evolved from mid-century responsibility charting in 1970; the primary source is cited in full below.
Where this is contested
No single originator exists: responsibility charting emerged from 1950s to 1970s management practice, with NASA-era responsibility assignment matrices, DuPont and the Goal Directed Project Management method all cited as sources, and the RACI initialism itself has no identifiable author.
- Format
- Checklist / audit
- Level
- Business unit · Team
- Best for
- Plan execution · Allocate resources
- Decision stage
- Plan · Execute
- Difficulty
- Introductory
- Time to apply
- Two to four hours in a workshop for a process of 10 to 25 tasks. The arguments it surfaces can take longer, and they are the point.
Plate · The model
The components
Responsible
The person or role that does the work to complete the task. Every task needs at least one R; more than one is workable but each addition dilutes ownership of the doing.
Signals of strength
Named roles or individuals rather than departments · Every row has at least one R · The R agrees, out loud, that they are the R
Accountable
The single name that answers for the outcome and signs it off. Exactly one per task, holding real authority over it; often senior to the R, though it need not be. This is the load-bearing letter.
Signals of strength
Exactly one A per row, without exception · The A has genuine authority over the task · No role holds the A for so many rows that accountability is nominal
Consulted
People whose input is sought before or during the work, through two-way communication, because what they know can change what is done. The C list is where charts quietly rot.
Signals of strength
Each C can say what their input would change · Consultation happens before the work, rather than as review theatre after it · The list is short enough that the task can still move
Informed
People kept up to date on progress or completion through one-way communication. They need to know; they do not get a vote, and the chart should make that explicit.
Signals of strength
Informing is one-way and after the fact · Nobody on the I list believes they are really a C · A defined channel exists for informing, so it actually happens
When it earns its keep
- Work keeps falling into the gaps between functions at a handover, and each side sincerely believes the dropped task belonged to the other.
- A reorganisation or merger has left role boundaries ambiguous and the same arguments about ownership keep recurring.
- A recurring process has accumulated too many people: every decision waits on a crowd of reviewers and nobody can say which of them actually matters.
- A cross-departmental project is starting and you want the ownership arguments in week zero, on paper, rather than in month three, in an incident review.
And when it doesn't
- A small, stable team where everyone already knows who does what. The chart adds administration without adding information.
- The real problem is capability or trust. A matrix records who should act; it cannot make an unwilling or unable person act, and charting around them postpones the real conversation.
- A fast-moving product team with genuinely collective ownership. Parcelling a shared outcome into individually owned fragments fights the operating model that makes the team work.
- What you actually need is clarity on a few high-stakes decisions. Decision-rights tools such as Bain's RAPID are built for that job; RACI charts tasks, and stretches badly over one-off strategic choices.
How to run it
Before starting, gather the inputs the analysis depends on:
- A defined process broken into concrete tasks or decisions, at the level of granularity where the confusion actually lives.
- The real cast list: every role or named person touching the process, including those outside the immediate team.
- Honest knowledge of how the work is done today, from the people who do it, since charting the official version merely laminates the fiction.
- An owner with the authority to settle the arguments the chart will surface, because it will surface them.
- 1
List the tasks before the people
Rows are tasks or decisions, typically 10 to 25 of them, pitched where the confusion lives. Rows as coarse as 'manage quality' assign nothing anyone can act on; rows as fine as individual emails drown the signal. Columns are the roles or named individuals involved.
- 2
Assign the Accountable, exactly one per row
The A is the single name that answers for the outcome and holds the yes or no. This is the rule that gives the tool its force: shared accountability reliably becomes no accountability. If a workshop cannot agree the single A for a row, it has located the organisation's actual problem, and that discovery is worth more than the finished chart.
- 3
Assign Responsible, then prune Consulted and Informed
The R does the work; at least one per row, and every additional R dilutes. Then apply the tests: a C's input must be capable of changing the work, and must be sought before it is done; an I merely needs to know afterwards. The C column is where charts bloat, because a C is handed out to avoid giving offence.
- 4
Read the chart for patterns
Scan vertically as well as horizontally. A column dense with A's is a bottleneck; a column dense with C's is a brake on everything; a column with nothing in it is a passenger; a row with no R is work that only happens when somebody remembers.
- 5
Test against a live failure
Walk a recent dropped handover or stalled decision through the draft. If the chart would not have prevented it, the rows are wrong or the assignments are aspirational. Fix the disagreements in the room; a chart published with unresolved arguments embedded in it settles nothing.
- 6
Choose your variant deliberately, and keep the chart alive
RASCI adds S for Supportive, resources who assist the R, useful where delivery genuinely depends on helpers. RAPID, Bain's decision-rights tool (Recommend, Agree, Perform, Input, Decide), is the sibling built for decisions rather than tasks. Whichever you use, review it when the process or organisation changes; a stale chart is a record of an old argument.
Reading the result
A one-page grid showing, for each task in a process, who does the work, who answers for it, who must be consulted and who is merely told, plus a list of the ownership disputes the charting exercise forced into the open.
- Read the columns before the rows. The diagnosis is in the patterns: an overloaded A column is a bottleneck, a heavy C column is a brake, an empty column is a passenger.
- Treat every argument the chart provokes as the tool working. The disagreements over who holds the A already existed; the chart makes them discussable before the failure rather than after it.
- Treat more than one A on any row as an error state rather than a compromise, and a row with no R as work that is currently happening by luck.
A worked example
A volume housebuilder charts its plot-completion handover
A UK volume housebuilder completing around 900 plots a year across 14 sites has a recurring failure at the handover from build to sales and customer care. Plots reach legal completion day with snags open, warranty certificates missing, and buyers occasionally arriving to no keys. Site teams blame sales progression; sales blames site. The regional managing director commissions a RACI workshop on the twelve tasks between build-complete and key handover.
- Responsible
- Mapping the R's exposed both duplication and a void. 'Close out the snag list' was claimed by the site manager and the customer care technician, who had been quietly redoing each other's work; the chart split it at completion date, site manager before, customer care after. Worse, no row-holder existed for confirming utility connections: it had no R at all and happened only when someone remembered, which explained two of the year's ugliest move-in days.
- Accountable
- The one-A rule bit hardest on 'release plot for legal completion'. The build director and sales director both believed they held it, and in practice the release went to whoever pressed hardest that week, which is how unfinished plots completed. After a two-hour argument, the single A landed on the head of customer care, armed with a documented release checklist neither director can override informally. That argument was the most valuable output of the exercise.
- Consulted
- Sales progressors had been consulting everyone on every plot, solicitors, site, commercial, warranty inspectors, which consumed their week and slowed routine completions. The chart cut the C list on standard plots to the site manager alone, reserving full consultation for flagged plots. The warranty inspector was formalised as a C on the pre-completion inspection row, rather than being discovered as a blocker at the gate.
- Informed
- Buyers' solicitors and the selling advisor moved to I, fed by an automated completion-pack notification at release. Previously they had chased progress by phone, effectively promoting themselves to C several times a day. Making them formally Informed, with a reliable channel, removed most of the inbound noise that had made the progressors' overload look like a staffing problem.
The read. Within two quarters, completion-day failures fell from roughly one plot in five to about one in twenty, and most of the gain traces to a single cell of the chart, the lone A on plot release. The honest limit: the matrix did nothing for build quality itself. Snag volumes were unchanged; they were simply caught before completion rather than after the buyer moved in. The chart now gets reviewed at each regional restructure, because the first version quietly expired when two roles merged.
Pitfalls
- Allowing two Accountables on a row as a diplomatic compromise. It reproduces on paper exactly the ambiguity the chart exists to remove.
- Writing rows so coarse ('manage quality', 'own the customer') that the chart assigns nothing anyone could actually do or check.
- Consultation inflation: handing out C's to avoid offence, then wondering why every task waits on five opinions. Each C must be able to say what their input changes.
- Assuming the R and A distinction is self-evident. Everyday English blurs responsible and accountable, and some organisations swap the meanings; without an agreed definition the same chart reads differently across the same table.
- Filing the chart. A RACI that is invisible at the point of work, and unrevised when the process or organisation changes, is a record of an old argument rather than a working tool.
What the critics say
Decision-rights researchers argue that task-level responsibility charting misses what actually stalls organisations, which is decisions. Rogers and Blenko developed RAPID at Bain on the finding that clarity about decision roles, rather than task roles, is what separates high-performing organisations; for high-stakes choices a decision-rights tool does work a RACI cannot.
Rogers, P. and Blenko, M. W. (2006) 'Who Has the D? How Clear Decision Roles Enhance Organizational Performance', Harvard Business Review, 84(1), January 2006.
Agile practitioners charge RACI with institutionalising hand-offs and diffusing ownership: a cross-functional team collectively accountable for an outcome, as in Scrum's whole-team accountability, has no use for a matrix that parcels the outcome into individually owned fragments, and imposing one can recreate the silos the team structure was built to remove.
The matrix records intended behaviour, not actual behaviour, and there is little empirical research testing whether RACI charting improves outcomes. Its authority is customary, carried by decades of project management practice rather than by evidence, and charts are routinely produced, filed and never consulted again.
Work it through
Pick one task or decision and use the four blocks to record who is Responsible, Accountable (one name only), Consulted and Informed. Your entries persist for this browser session and can be copied out as Markdown or printed.
0 of 4 blocks filled
Run it as a workshop
Bring the actual task or process list into the room already drafted, ten to twenty-five rows is the workable range; building the task list live eats the whole session. Put one unbreakable rule on the wall before starting: exactly one Accountable per row.
Running order
| Confirm the task list and its granularity | 15 min |
| Assign Accountable first, one name per row, no exceptions | 30 min |
| Assign Responsible | 20 min |
| Assign Consulted and Informed, sparingly | 25 min |
| Surface and resolve the disagreements the chart provokes | 30 min |
| Total | 2h |
You will need
- The pre-drafted task or process list, one row per task
- A RACI grid, printed large, task rows against four role columns
- Sticky dots or initials for each cell
- A visible reminder of the one-Accountable rule
Traps to avoid
- Allowing two Accountables on a row as a diplomatic compromise. It reproduces on paper exactly the ambiguity the chart exists to remove.
- Writing rows too coarse to assign anything real, 'manage quality', 'own the customer'. Push for task-level specificity or the chart assigns nothing anyone could actually do or check.
- Consultation inflation: handing out C's to avoid offending anyone, until every task waits on five opinions and none of them can say what their input would change.
- Treating an argument the chart provokes as a failure of the workshop. That disagreement already existed; the chart has simply made it discussable, which is the tool working, not breaking.
Sources and further reading
- Project Management Institute (2017) A Guide to the Project Management Body of Knowledge (PMBOK Guide). 6th edn. Newtown Square, PA: PMI. (Documents the responsibility assignment matrix and RACI as standard practice.)
- Andersen, E. S., Grude, K. V. and Haug, T. (1987) Goal Directed Project Management. London: Kogan Page. (The GDPM tradition of responsibility charting.)
- Rogers, P. and Blenko, M. W. (2006) 'Who Has the D? How Clear Decision Roles Enhance Organizational Performance', Harvard Business Review, 84(1), January 2006. ↗
- Wikipedia: Responsibility assignment matrix (survey of the tool's history and variants). ↗