---
title: "Design an RFP responsibility matrix that exposes gaps"
description: "Map bid outputs to accountable roles, contributors and scoped decision rights. Test missing owners, conflicting duties and unaccepted handoffs before use."
canonical: "https://zephior.com/insights/design-an-rfp-responsibility-matrix"
last-updated: 2026-09-06
---

# Design an RFP responsibility matrix that exposes gaps

> Map bid outputs to accountable roles, contributors and scoped decision rights. Test missing owners, conflicting duties and unaccepted handoffs before use.

By [Tony Kim](https://zephior.com/authors/tony-kim). Published 2026-09-06; updated 2026-09-06. 15 minute read.

## Definition

An RFP responsibility matrix connects each controlled bid output or decision to a production lead, one accountable role, supporting and advisory roles, and its receiving interface. A companion authority record specifies who may make which decision under which mandate. The result is a testable operating design, not evidence that anyone has completed, approved or submitted the offer.

## Problem

A spreadsheet gives both the technical director and the commercial director an A against approve the bid. It gives the bid manager an R everywhere else. Every row contains a name, yet nobody can say who receives a failed review, whether commercial approval covers the technical promise, or who keeps the portal receipt. Adding more names makes the sheet denser without settling those questions.

## Point of view

Build the matrix around inspectable outputs and test its interfaces before treating it as active. This guide owns the whole-bid allocation design. The work-breakdown guide defines the outputs; the answer-owner guide selects suitable people; specialist approval records establish actual mandates. The fictional Norchester Instruments case tests the matrix without prescribing a universal corporate hierarchy. Sources were checked on 6 September 2026.

## Use outputs another role can accept, not department names

A useful row ends in something inspectable: a technical-position decision, a populated price model, a review finding, an assembled answer or a submission receipt. Engineering and finance belong in the role model, not in the output column. Otherwise the sheet cannot show whether two departments are working on the same object or whether an object has disappeared between them.

NASA’s 2021 WBS Handbook, section 4.2, connects defined work products with the organization through a responsibility assignment matrix. It warns against allowing the organization chart to determine the work breakdown. The bid-specific checks here adapt that project-management distinction; they are not NASA procurement requirements.

Import the work-package scope rather than rewriting it to suit whoever is available. Retain the question, lot, bidder entity, version and output identifier. Include internal decisions that downstream work consumes, even when the buyer never sees them. A decision to permit a particular commercial position and the page that describes it are separate objects with different completion tests.

Split a row when its parts have different authority boundaries, recipients or acceptance criteria. Do not split every sentence. Three specialists can contribute to one integrated answer while its owner remains clear. Conversely, a broad row called approve answer can hide technical validation, pricing consent and permission to release. Those decisions should not disappear into one letter.

Keep inactive alternatives visible with their activation conditions. A partner route that has not been selected need not carry active production assignments, but it must not look ready simply because a previous tender used that partner. The matrix should say which branch was tested and what must be assigned if the choice changes.

**Fields that make a responsibility row inspectable**

| Field | What to record | Failure exposed |
| --- | --- | --- |
| Output identity | Stable identifier, scope, entity, lot and active branch | Two teams appear to own one thing but mean different scopes |
| Production and accountability | One production lead and one accountable role under the stated legend | Nobody integrates the contributors or answers for completion |
| Supporting and advisory work | Specific contribution or question, not an unlimited department label | A requested deliverable is mistaken for optional advice |
| Decision authority | Decision type, mandate reference, limits and required concurrences | A coordinator is expected to approve beyond their authority |
| Receiving interface | Recipient, expected state, acceptance evidence and rejection route | An output is sent but no one owns what happens next |
| Activation evidence | Confirmed people, assignment acceptance and unresolved restrictions | A well-designed role exists only on paper |

## Specify what a letter does and what it cannot authorize

PM² Guide 3.1, section 4.12, distinguishes R, the work lead; A, answerable for completion; S, hands-on support; C, advice through two-way communication; and I, progress information. Its convention uses one R and one A per task. Use a declared legend rather than assuming every organization interprets RACI identically.

This method records approval powers separately. On a decision row, the production lead prepares and records the decision process; the authorized decision maker must still make the decision. A row’s accountable role must have a credible means of arranging completion, but being answerable for an answer does not give that role the power to change prices, waive a policy, bind a partner or submit the offer. Where the same person also holds one of those powers, record the separate mandate and its scope.

The UK government’s GovS 002, sections 4.3 and 4.4.1, links decisions to the governance framework and defines roles through activities, outputs and accountability. It applies in its government context. For a bidder, the useful question is whether the proposed allocation agrees with its own governing authority, rather than merely with a job title.

An S assignment needs a deliverable. If a service engineer must supply a capacity calculation, write that obligation and its destination; C alone would imply advice rather than production under this legend. Consultation also needs a defined question and a way to close or escalate objections. Silence is not affirmative consent, and a distribution-list entry proves neither reading nor acceptance.

The one-A convention is a design choice for each bounded output, not a claim that all legal or executive accountability can be reduced to one person. Higher-level responsibility remains. Where governance requires a collective body, name that body, its operating representative, decision rule and required membership or quorum. Do not appoint an individual with fictional powers to satisfy a spreadsheet check.

## Six accountable roles still leave a broken handoff

Norchester Instruments is fictional. Its team initially puts the technical and commercial authorities together as A against approve bid. For this example, its verified internal mandates give those authorities different decision scopes and reserve package release to a separate role. The team replaces the broad row with the six outputs below. The role names illustrate those assumed mandates; they do not establish a recommended hierarchy for another business.

T1 settles the technical boundary. P1 settles the commercial position for that boundary. A1 assembles their compatible results into the response. Q1 records the independent review result. D1 decides whether the exact package may be released. U1 produces evidence of the authorized transmission. Support and consultation are recorded in the full matrix; the compact view focuses on the production, accountability and receiving interfaces.

At the design review, all six outputs have exactly one A and one R. Five receiving interfaces have confirmed acceptance rules. The missing one is Q1 to D1: the release authority expects a clean recommendation, while the quality lead expects to send the report even if findings remain unresolved. Neither has accepted responsibility for deciding how that unresolved report enters the release process.

The matrix is fully populated on the selected ownership test, but only five of six mandatory interfaces are agreed, or 83.3% rounded to one decimal. This is an interface-design count, not a completion percentage or probability of successful submission. T1 and P1 could still be unperformed, and the six outputs are only the declared slice of the offer.

**Norchester: proposed allocation for six bounded outputs**

| Output | R: production lead | A: accountable role | Decision boundary | Required receiver |
| --- | --- | --- | --- | --- |
| T1 technical-position decision | Engineering analyst | Technical authority | Specified technical scope only | A1 integrator: interface agreed |
| P1 commercial-position decision | Pricing analyst | Commercial authority | Defined commercial package only | A1 integrator: interface agreed |
| A1 assembled answer | Response writer | Answer integrator | Readiness for review, not new commitments | Q1 review lead: interface agreed |
| Q1 review report | Independent reviewer | Quality lead | Findings and review conclusion, not release | D1 release authority: unresolved receiving rule |
| D1 package-release decision | Release coordinator | Release authority | Exact package and all required valid decisions | U1 operator: interface agreed |
| U1 transmission receipt record | Portal operator | Submission lead | Transmission only after applicable release authority | Archive custodian: interface agreed |

## Define how a rejected output moves forward

Norchester’s repair is an agreed Q1 receiving rule. The release authority accepts the report into its decision process whether the review conclusion is clean or adverse. The report identifies the exact reviewed version and unresolved findings; each finding has an assigned resolution or referral route. Accepting receipt of an adverse report does not accept its risk or authorize release.

The authority may return the package for correction, request a competent decision on a reserved matter or refuse release. It cannot override another role’s required technical or commercial decision merely because it receives Q1. If the actual governance permits a particular override, that power and its limits require separate evidence. The matrix should preserve the distinction between moving the issue and deciding it.

After the parties confirm this rule, six of six interfaces in the selected slice are agreed. Record the change as a design repair, not six completed outputs. Q1 may still reveal a defect later. U1 also remains separate: an operator’s account access and ability to upload do not prove that D1 authorized the file the operator selected.

Test the negative path on other interfaces. What happens when a contribution arrives for the wrong lot, a required approval is refused, or a receiver rejects the version? An interface that describes only successful delivery is incomplete. The sending role retains the work or follows the agreed referral route until another authorized role accepts the next responsibility; it does not mark itself finished solely because it sent an email.

## Check across rows as well as across columns

The GAO’s 2025 Green Book, paragraphs 10.21 to 10.23, addresses incompatible duties and alternative controls where segregation is impractical. It is a US federal internal-control framework. Applying that risk question to bid review does not make every shared role prohibited or establish that any particular substitute control is sufficient.

Bind role identifiers to the proposed people and inspect the result. Two role labels can refer to the same individual. If Norchester requires Q1 to be independent of the author of A1, assigning both roles to one person fails that rule even though every row has a different title. A second account, a later timestamp or a deputy who reports the same unexamined conclusion does not create independent review.

Also inspect conditional dependencies. If technical approval requires an already approved price and commercial approval requires an already approved technical position, the matrix contains a circular prerequisite. More reminders will not settle it. The authorized process owner must distinguish preliminary analyses from final decisions, define a permitted joint decision route, or refer the conflict. An assistant must not weaken the approval conditions to break the cycle.

Look for one role appearing as the production lead on many outputs. That is a workload warning, not proof of overload: row counts contain no duration or availability evidence. Send the affected assignments to the separate capacity and scheduling checks. Likewise, a matrix does not need a recurring meeting in every cell. A meeting may support a decision, but its cadence does not define who may make that decision.

**Static tests before activating the matrix**

| Test | Inspectable failure | Required disposition |
| --- | --- | --- |
| Coverage | An active output or required decision has no row | Restore the missing allocation with its scope |
| Uniqueness | No A, several unexplained As, or no production lead | Resolve the boundary or governing collective rule |
| Authority | The requested decision exceeds the referenced mandate | Obtain the proper decision route; do not rename the role |
| Independence | The same person holds duties that the stated control separates | Assign an eligible reviewer or seek an authorized alternative |
| Interface | A mandatory receiver or rejection rule is missing | Agree the next responsibility before relying on the handoff |
| Dependency | Final decisions wait on each other without an authorized resolution | Repair the decision process and retest affected rows |

## Issue the matrix only for the assignments that were accepted

The handoff contains the output-to-role matrix, legend, authority references, people bindings and interface-test record. Give the package a revision and state who authorized its use. Proposed, accepted and active assignments need distinct states. Where a gap affects only one separable branch, the authorized manager can define bounded work elsewhere; the existence of a gap does not itself authorize either a universal stop or unrestricted continuation.

Role holders should confirm the precise duty they accept. A partner’s liaison cannot be assigned authority over the partner’s commitments merely because they attend bid meetings. Obtain the counterparty’s authorized role confirmation through the permitted route, retain the legal-entity boundary and record the receiver on the bidder’s side. Acceptance of coordination work is narrower than acceptance of the proposed contractual obligation.

Revalidate affected rows when an output splits, the bidder entity changes, a role holder leaves, a delegation expires, an independence conflict emerges or a new partner branch becomes active. Preserve the previous revision and the reason for the change. A substitute may receive the production task without receiving the original holder’s reserved powers. Reference the actual replacement mandate rather than copying the old person’s approval cells.

An assistant can extract candidate outputs from authorized records, propose allocations for review and run reproducible gap, identity and dependency checks. It must keep unsupported assignments unresolved. It cannot appoint staff, obtain access by changing permissions, accept a role on someone’s behalf, approve an exception, contact a partner or release an offer without separate authorization. Public guidance describes the method; the operational matrix may contain personal and commercially sensitive information and needs controlled access.

## Useful outcomes

- Each active output has a bounded purpose and one accountable role.
- Producing work, advising and authorizing a decision remain distinguishable.
- Mandatory joint approvals survive the simplification of the matrix.
- Role assignments reveal person-level conflicts before activation.
- Every mandatory handoff has a recipient and an acceptance rule.
- The issued record identifies unresolved gaps and the limits of permitted work.

## Workflow

1. **Import the controlled outputs.** Take the current work-package identifiers, required decisions, applicability branches and receiving artifacts. Add omitted control outputs such as a review report or receipt record without recreating the whole breakdown.
2. **Agree what the letters mean.** Publish one legend, identify the production lead and accountable role for each row, and distinguish hands-on support from consultation. Keep formal decision powers in explicit fields.
3. **Resolve compound decisions.** Separate technical acceptance, commercial authorization and assembled-answer readiness where they have different tests or mandates. Preserve required joint concurrence in the actual decision rule.
4. **Test roles and interfaces.** Check missing and duplicate assignments, incompatible duties, authority boundaries, required receivers and circular prerequisites. Bind roles to people to expose conflicts hidden by job titles.
5. **Obtain bounded acceptance.** Have the authorized manager, role holders and receiving parties confirm their exact scope, permitted access and handoff conditions. A proposed name is not an accepted assignment.
6. **Issue the tested revision.** Preserve the matrix, authority references, person bindings, interface results and open exceptions together. Identify which work may start and what must remain held until the design is repaired.

## Key decisions

- Is this row one output or several decisions disguised as one?
- Which role answers for completion, and which mandate permits the requested decision?
- Does a joint decision require a constituted body or several explicit concurrences?
- Can the named people satisfy the required separation of duties?
- Who accepts each handoff, including a rejection or unresolved issue?
- Which changes invalidate the current assignment design?

## Risks

- Treating an A cell as unlimited approval authority
- Listing a working contributor only as someone to consult
- Removing a mandatory approval to make a row contain one A
- Hiding self-review behind two different role labels
- Assigning work to a partner without its authorized acceptance
- Calling a populated matrix proof of bid readiness

## Metrics

- Active outputs lacking an accountable role
- Compound rows with unresolved decision boundaries
- Mandatory handoffs without accepted receiving responsibility
- Person-level conflicts against stated independence rules
- Assignments awaiting mandate or role-holder confirmation
- Changed interfaces awaiting matrix revalidation

## Frequently asked questions

### Does A always mean the person who signs the offer?

No. Use the declared legend and the actual delegation. Accountability for completing an output, approval of its contents, legal signature and transmission can belong to different roles.

### Can several required approvals coexist with one A?

Yes. Preserve every required concurrence in the decision rule. Split distinct decision objects or identify the properly constituted accountable body; do not remove a required approval just to simplify a row.

### Is a subject-matter expert a contributor or someone to consult?

It depends on the work requested. Under this guide’s legend, producing a calculation is supporting work with an output, while advising on a defined question is consultation. State the obligation explicitly.

### What does six of six accountable rows prove?

Only that the declared rows pass that assignment-count test. Norchester still has a missing receiving rule before repair. The count proves neither completed work nor valid approval.

### Can the matrix replace the bid schedule?

No. It can expose dependencies and concentrated assignments, but it does not establish durations, calendars or capacity. Pass those inputs to scheduling and keep meeting cadence outside the allocation artifact.


## Primary sources

- [NASA WBS Handbook, 2021, section 4.2: work assignment](https://www.nasa.gov/wp-content/uploads/2023/08/nasa-work-breakdown-structure-handbook.pdf), NASA
- [PM² Guide 3.1, section 4.12: RASCI](https://op.europa.eu/en/publication-detail/-/publication/97cc2f12-c648-11ee-95d9-01aa75ed71a1/language-en), European Commission
- [GovS 002, sections 4.3 and 4.4.1: decisions and roles](https://projectdelivery.gov.uk/library-products/government-functional-standard-govs-002-project-delivery/), UK Government
- [Green Book 2025, paragraphs 10.21 to 10.23: incompatible duties](https://www.gao.gov/assets/gao-25-107721.pdf), US Government Accountability Office


## Related articles

- [Turn RFP questions into a controlled work breakdown](https://zephior.com/insights/build-a-work-breakdown-from-rfp-questions)
- [Choose an RFP answer owner across teams](https://zephior.com/insights/assign-rfp-owners-by-evidence)
- [Who is allowed to approve each kind of proposal evidence?](https://zephior.com/insights/assign-ownership-for-proposal-evidence)
- [Set bid escalation rules before the decision window closes](https://zephior.com/insights/define-escalation-rules-for-a-bid)