1. The Grid
16 decisions, 8 roles. A is accountable — exactly one per row, enforced at build time. R does the work. C is consulted and can change the outcome. I is told. A dash means no role at all.
| Decision | PD Tyrrell | CMO Okoye | CSO Vasquez | REG Raghunathan | CFO Thornbury | MA Barrington | CQO Solberg | DMC Vartanian | Note |
|---|---|---|---|---|---|---|---|---|---|
| Authorize a stage tranche | R | C | C | C | A | I | I | - | ⚠ The PD proposes; the CFO is accountable for releasing money. |
| Approve the program plan and baselines | A | C | C | C | C | C | I | - | One of only TWO rows where the PD is accountable. |
| Approve a protocol or protocol amendment | C | A | C | C | I | I | C | I | Medical judgment. The PD coordinates and does not decide. |
| Set or revise the target product profile | C | C | A | C | C | R | I | - | Scientific ownership, with market access driving the commercial attributes. |
| Approve the statistical analysis plan | I | C | R | C | - | - | C | A | ⚠⚠ Accountability sits with the DMC chair as the independent party. The sponsor cannot be accountable for the plan that judges its own trial. |
| Authorize database lock and unblinding | R | C | C | I | - | - | C | A | Same reasoning. The sequence is the integrity of the trial. |
| Approve a submission to the agency | R | C | C | A | I | I | C | - | Regulatory owns what is said to the agency, and carries the commitments made. |
| Accept or decline an expedited designation | C | C | C | A | I | C | - | - | A written argument to the agency about our own product. |
| Approve a change request | R | C | C | C | A | C | I | - | Scope changes cost money; the CFO is accountable for the ceiling. |
| Draw against contingency | R | I | I | I | A | I | I | - | Proposed by the program, released by finance, with the rejected alternative recorded. |
| Accept a quality deviation or CAPA | I | C | - | C | - | - | A | - | ⚠ The CQO reports to the CEO, not to this program. Deliberate. |
| Stop the trial for safety | I | C | I | I | - | - | C | A | ⚠⚠ Outside the Committee's authority entirely. The program cannot manage the body that might stop it. |
| Approve pricing and the rebate ceiling | C | I | - | I | A | R | - | - | Recommended by market access, accountable to finance. |
| Approve the launch plan | C | C | - | C | C | A | C | - | Commercial ownership, with regulatory consulted on promotional boundaries. |
| Accept a post-marketing commitment | R | C | C | A | I | I | C | - | It is a regulatory commitment and it outlives the program. |
| Close the program | A | C | I | C | C | C | C | - | The second and last of the PD-accountable rows. |
| Who | Role | |
|---|---|---|
| PD | C. Tyrrell | Program Director |
| CMO | Dr. A. Okoye | Chief Medical Officer / sponsor |
| CSO | Dr. H. Vasquez | Chief Scientific Officer |
| REG | Dr. P. Raghunathan | SVP Regulatory Affairs |
| CFO | M. Thornbury | Chief Financial Officer |
| MA | J. Barrington | SVP Market Access |
| CQO | Dr. I. Solberg | Chief Quality Officer |
| DMC | Prof. E. Vartanian | Data Monitoring Committee chair |
2. What the Grid Says About the Job
The Program Director is accountable for 2 of 16 decisions: approving the plan and its baselines, and closing the program. Responsible on nine. Consulted or informed on the rest.
That is not a weakness in this grid; it is an accurate description of the discipline. Protocol decisions belong to medicine, money to finance, submissions to regulatory, safety to a committee that reports to nobody here. What the Program Director owns is that each decision is taken by the right person, on the right evidence, at the right time, and recorded.
The two rows where accountability does sit with the PD are the two that are about the program itself rather than about the product — how it will be run, and when it ends.
The R column is the mirror image and is worth reading alongside it. The Program Director is Responsible on nine rows — proposing the tranche, running the change request, preparing the submission, driving the contingency draw. Doing the work of a decision somebody else owns is most of the job, and a grid that showed the PD accountable everywhere would have described a role that does not exist in a matrix organization.
The fact base asserts this. If the PD were accountable for more than four rows, the build would fail on the grounds that the grid is describing a line-management role rather than a program-management one.
3. Four Decisions Accountable Outside the Program
4 rows place accountability with someone the program cannot direct. Every one is deliberate.
| Decision | Accountable | Why not the program |
|---|---|---|
| Approve the statistical analysis plan | DMC chair | ⚠ The sponsor cannot be accountable for the plan that judges its own trial. The analysis order decides what may be claimed. |
| Authorize database lock and unblinding | DMC chair | The sequence — SAP signed, then lock, then unblind — is the integrity of the trial. Reverse any two and the result becomes unfalsifiable. |
| Accept a quality deviation or CAPA | Chief Quality Officer | Reports to the CEO, not to this program. An audit function accountable to the party it audits is not independent. |
| Stop the trial for safety | DMC chair | ⚠⚠ Entirely outside the Committee's authority. The program cannot manage the body that might stop it. |
These four are the structural counterweight to everything else in the suite. A governance model in which every decision is accountable inside the program is one where nothing can stop the program — and a program that cannot be stopped is not being governed, it is being administered.
4. The Rules That Keep It Honest
| Rule | Here | Why it matters |
|---|---|---|
| Exactly one A per row | Enforced at build time | ⚠ Two names in the A column means nobody is accountable. Shared accountability is the most common and most damaging RACI defect, and it usually reflects a decision nobody wanted to make about who decides. |
| R and A may be the same person, and usually are not | Separated on 13 of 16 rows here | Responsible does the work; Accountable answers for it. Where they separate, the artifact is describing real delegation rather than a job description. |
| C means their view can change the outcome | Used sparingly | ⚠ If a party is consulted and the answer never changes, they are Informed. Inflating C is how a RACI becomes a courtesy list and stops being read. |
| I means told after the fact | Not a lesser status | Some of the most senior people on this grid are I on rows they have no business deciding. |
| A dash means no role at all | Used deliberately | A grid with no blanks has not been thought about. Not everyone belongs on every row. |
Consulting somebody is expensive — it means waiting for their view and being willing to change course because of it. Marking a party C when the answer never changes costs the same time and produces a document nobody trusts, because the grid now describes a courtesy rather than an authority.
The test is simple: if their view could not change the outcome, they are I. Demoting a senior person from C to I is uncomfortable and it is what makes the remaining Cs mean something.
This grid is a subsidiary of the governance model and does not override it. Where the two disagree, the governance model wins — and the disagreement is itself a finding, because a RACI that has drifted from the charter it implements is describing how a program actually behaves rather than how it was authorized to.
5. When the Grid Does Not Answer
A RACI covers the decisions somebody thought of. Most of what a program actually decides is not on it, and the useful question is what happens then.
| Situation | What happens | Who resolves it |
|---|---|---|
| A decision that is not on the grid | It is mapped to the nearest row by type rather than by subject. A novel question about analysis order is an SAP decision; a novel question about money is a tranche decision. ⚠ The grid is a set of decision classes, not a list of questions. | Program Director proposes the mapping; the accountable party for that class confirms it. |
| R and A disagree | A decides. That is what accountable means, and it is the situation the single-A rule exists for. ⚠ The disagreement is minuted, because a decision taken over a documented objection is a different fact from an unopposed one. | No escalation needed. The grid already answered it. |
| Two parties both believe they are A | ⚠⚠ A defect in the grid, not a dispute between people. The row is re-drafted and taken to the Committee — not settled informally between the two claimants. | Development Committee, as owner of the governance model. |
| The accountable party is unavailable | Named deputies exist for every A. ⚠ Accountability delegates to a person, never to a function — "Regulatory Affairs" cannot be accountable at the moment a decision is missed. | The deputy, on the record, with the same authority. |
| The grid contradicts the charter | The charter wins, and the contradiction is itself reported. | ⚠ A RACI that has drifted from the document it implements is describing how the program behaves rather than how it was authorized to. |
When two people both believe they are accountable, the instinct is to let them work it out. That produces a private accommodation between two individuals about a decision the organization has an interest in — and it produces no record. Treating it as a defect in the grid rather than as a disagreement between people sends it to the body that can actually change the answer.
It is also the reason the single-A rule is enforced at build time rather than left to review. A grid with two As in a row does not announce itself; it simply sits there until the day it matters.
Maintenance is deliberately minimal. The grid is re-confirmed at each gate, alongside the governance model and the baselines, and changed only by the Committee. It is not adjusted when people change roles — the roles are the rows, and a new Chief Financial Officer inherits the CFO column intact. A RACI that is updated whenever the organization moves is tracking the organization; this one tracks the decisions.