How to build a RACI

This guide helps an SME build, in a small number of working sessions, a RACI matrix the organisation will actually use. The goal is not to produce a formal table for an archive. The goal is to make responsibility visible on the activities where ambiguity creates delay, rework or escalation.

What this guide is for

A useful RACI states, for the critical activities of a flow, who performs the work, who is ultimately accountable for the result, who must be consulted before completion and who only needs to be informed afterwards.

The “without creating bureaucracy” part matters. Most RACI matrices die because they are filled on imagined work instead of real work, loaded with political consultation, and never used when the flow gets stuck.

When to use it

Use it when the same misunderstandings repeat: “I thought you were handling it”, “I did not know you needed my approval”, “I was waiting for the OK”.

Use it when authorisation waits dominate lead time.

Use it when the founder wants to delegate structurally, with declared boundaries rather than informal trust.

Use it before extraordinary operations or growth plans. An explicit responsibility map reduces perceived key person risk because it shows that work is not entirely coordinated through the founder.

Before starting

You need a map of the flow, even a rough one. If you do not have it, start with process mapping. RACI is built by flow, not by organisation chart. The rows are real activities, not abstract functions.

Invite the people who actually touch the activities, not only their managers. Managers tend to describe the nominal process. The RACI must be built on the real process.

Make one rule non-negotiable before the session begins: one Accountable per activity. Two Accountables usually mean none. This rule will meet political resistance, so protect it early.

Operational sequence

Step 1 — Choose the right granularity.

From the process map, select 8 to 15 activities where misunderstanding can occur: approvals, handoffs, exception handling, customer communication, release decisions, changes after confirmation.

Do not map macro-areas such as “sales is responsible for revenue”. They decide nothing. Do not map every micro-action either. The right unit is the activity that can fall into the void.

Step 2 — Fill the matrix together, one row at a time.

Put the map on the table. For each activity ask four questions.

Who actually does this today? That is R.

Who answers if it goes wrong? That is A.

Who must give input before it is completed? That is C.

Who only needs to know afterwards? That is I.

Write what emerges, not what should be true. Divergences during the session are the value of the exercise: two people convinced they are Accountable for the same activity, an activity with no Accountable, or a role that believes it must be consulted while others see it as only informed. These are role gaps measured in real time.

Step 3 — Apply the single-A test.

Every row with a contested or missing Accountable must be closed in the session. Who is ultimately answerable?

A useful criterion is proximity to consequence: the Accountable should be the role that first suffers the consequence if the activity fails and has the lever to prevent it.

If the choice is impossible without the founder, the founder decides the assignment. But the founder should not take the A simply to keep peace. That recreates the dependency the matrix is meant to reduce.

Step 4 — Prune the Consulted column.

This is the step that separates an operating RACI from a bureaucratic one.

Every Consulted role can slow or block the activity. For each C, ask: what changes in the result if this role is not consulted before completion?

If the answer is “nothing, but they might be upset”, downgrade the role to Informed. Information after the fact is cheap. Consultation before the fact costs lead time.

A healthy RACI contains many empty cells. Empty space is information: this role is legitimately not involved in this activity.

Step 5 — Attach thresholds to authorising activities.

For authorisations — discounts, purchases, delivery exceptions, credit limits — the Accountable alone is not enough. The boundary must be defined.

Examples: discounts up to 8% are decided by the sales manager; above that, the founder is required. Delivery exceptions up to five days are handled by operations; beyond that, escalation is needed.

Without thresholds, delegation remains reversible and people keep waiting for approval “just to be safe”. Thresholds are where RACI touches decision rights.

Step 6 — Use it out loud for eight weeks.

A RACI does not merely describe reality. It commits the organisation. An agreement that is never invoked disappears.

For two months, whenever something gets stuck, the first question in review should be “who was Accountable?” before “who is to blame?” That difference matters. The first question checks the architecture. The second processes people.

After several cycles, the matrix should become less necessary because people internalise it.

Step 7 — Update by event, not by calendar.

A RACI becomes obsolete when roles, tools, volumes or decision boundaries change, not every six months by ritual.

A practical rule: whoever proposes an organisational change that affects the flow also brings the updated RACI row. Maintenance stays distributed and the document remains alive.

Expected output

The expected output is one page: 8 to 15 activities, exactly one Accountable for each, only necessary Consulted roles, explicit thresholds for authorising activities.

Two by-products may be equally valuable: a list of role gaps that surfaced during the session and the first written delegation thresholds in the company.

Mistakes to avoid

Do not fill the RACI at a desk from the nominal process.

Do not allow double Accountable for political comfort.

Do not fill the Consulted column to buy consensus at the cost of lead time.

Do not build the matrix and then never ask “who was Accountable?”

Do not forget thresholds, leaving delegation reversible.

Do not start with the RACI of the whole company. A surgical one-page RACI for one flow works better than a total map that dies as a project.

Example

A 22-person commercial company has chronic waiting in its order flow. Mapping shows that 60% of lead time is waiting for “OK”: discount approval, delivery exception approval and customer credit approval.

A one-hour RACI session with six people reveals the structure. Out of 11 critical activities, four have contested Accountability between the sales director and founder. Two have no declared Accountable, including post-confirmation changes, the main source of conflict for two years.

The company assigns one Accountable per row and writes three thresholds: discounts up to 8%, delivery exceptions up to five days and credit within a set limit are delegated without founder approval.

After two months of using the RACI out loud in weekly reviews, median order lead time falls by 40%. The founder discovers that interruptions for signatures and approvals were not “about an hour a day”, as estimated, but three. Two disappear.

MARTRO connection

In MARTRO’s reading, RACI is a detector disguised as a table. The group session exposes, in one hour, the ambiguity that delays and conflicts had been signalling for years.

The finished matrix matters, but the divergences exposed while building it matter even more. Step 5, the threshold step, is where process work becomes governance work. Declaring who can decide what, and within which boundary, is one of the first acts by which an SME becomes less dependent on the founder’s omnipresence.

Next step

If threshold discussions reveal that the real issue is decision authority, continue with decision rights.

If the session exposed large divergence between how roles see responsibility, continue with role gap.

For the conceptual background, read RACI in the Knowledge library.

Frequently asked questions

How many activities should a first RACI include? Usually 8 to 15. Enough to cover the critical flow, not enough to become unmanageable.

Can two people be Accountable? Avoid it. Two Accountables usually mean no clear final owner. If joint ownership is unavoidable, define a tie-breaker or decision threshold.

Should everyone be consulted? No. Consultation before the fact costs lead time. Inform people after the fact when their input does not change the result.

Does RACI replace decision rights? No. RACI clarifies responsibility for activities. Decision rights clarify authority over choices. They meet when an activity involves approval or authorisation.

How do we keep the RACI alive? Use it in reviews for eight weeks. When something is stuck, ask who was Accountable. Update rows when roles, tools or thresholds change.