How do decisions get made in a matrix organisations?

I was working with a regional operations director leading a supply chain consolidation programme across a pan European business. She had done everything right. The business case was solid. The decision rights were clear, with each of the five RAPID roles, Recommend, Agree, Perform, Input, and Decide, assigned to the right person. The sign offs came in. The decision was made in March. By June it had largely stopped happening. Some divisions were still running the old process. Others were waiting to see what everyone else did first. Nobody had challenged the decision openly. They had simply not acted on it, and there was no mechanism that required them to.

What went wrong was not the process. The process was followed. What went wrong was the assumption underneath it: that a signed off decision and a delivered one are the same thing. In a hierarchy, that assumption usually holds. The person with the authority to decide also has authority over the people who need to execute. In a matrix, they are often not the same at all, and the gap between them is where decisions tend to fail.

Decision rights are not the same as decision power

Consider two divisions running the same decision rights process. Same RAPID chart. Same named Decider. Same sign offs from the same functions.

In the first division, the decision holds. Implementation moves forward, and when problems come up, they get raised and resolved rather than quietly worked around. In the second, the decision is technically made and practically ignored. The chart looks identical in both cases. What is different is invisible to the chart: whether the people whose work the decision depends on were actually committed to it, or simply present for the meeting where it was announced.

It helps to be precise about what RAPID actually assigns. Recommend builds the case and gathers the data. Agree must formally sign off, within a defined scope, before the decision proceeds. Input is consulted for expertise and perspective but holds no veto. Decide makes the single binding call. Perform carries out the decision once it is made. RAPID is useful for clarifying who holds each of these five roles, which is why it has become the standard reference for decision rights. What it does not address is whether the people holding Agree or Input have any practical obligation to act once Decide has spoken. The framework assumes that being correctly named in a role is the same as being bound by the outcome. In a matrix, it often is not.

This is the distinction worth naming clearly. Decision rights describe who holds formal authority over a decision. Decision power describes who has the practical ability, formal or not, to make that decision real or to quietly let it die. In a hierarchy, the two tend to converge, because the person with formal authority usually also controls the people needed to deliver. In a matrix, they routinely diverge. The named Decider often has no reporting authority over a number of the people whose behaviour determines whether the decision actually happens.

This is also why a lot of advice on matrix decision making falls short in practice. It answers the rights question well and stops there, as though clarity of process were the same thing as commitment to outcome. Research on matrix organisations published in MIT Sloan Management Review found that the failures of most matrix structures are caused as much by the speed and follow through of decisions as by the contest over who holds the rights to make them in the first place. Clear rights are necessary. They are not sufficient.

This article looks at what closes that gap, drawing on the four capabilities explored elsewhere in this series. The foundation article, How do successful leaders create commitment in matrix organisations?, introduced influence, networks, convening, and accountability as the four capabilities that make lateral leadership possible. The other deep dives cover what a matrix organisation is and why companies adopt one, how to build influence without authority, how to build a powerful internal network, how to convene people who don’t report to you, and how to create accountability without authority. Decision making is where all four of those capabilities get tested at once, because a decision is only ever as strong as the commitment, reach, ownership, and follow through around it.

Why decisions fail in matrix organisations

A Decider with no authority over key contributors: The RAPID process names a single Decider correctly, but several of the people whose Input or Agree was sought report to someone else entirely and feel no particular obligation to act once the decision is made. Their participation in the process was real. Their accountability to its outcome was not.

A role filled by someone who cannot actually speak for their function:  Sometimes the right person sits in the right RAPID seat, and the decision still does not hold, because that person was only ever able to represent part of their function rather than commit the whole of it. They agreed personally. Their division did not.

Treating agreement as commitment: A decision is signed off in the room, and everyone present nods. This is the same gap explored in the previous article in this series on accountability. A decision that exists only as a record of agreement, rather than as a set of specific commitments owned by named people, has no mechanism to survive contact with competing priorities once the meeting ends.

The silent veto:  Someone with no formal authority to block a decision can still prevent it from happening, simply by not prioritising it. They never say no. They never escalate. They quietly let other work take precedence, and because no formal mechanism required their cooperation, nothing forces the conflict into the open. The decision dies of inattention rather than disagreement.

Treating a complicated decision process for a complex problem: RAPID and similar frameworks work well for decisions where the right answer can be found through analysis and expert input, a complicated problem in the sense used by Snowden and Boone’s work on decision making, where cause and effect are knowable in advance. A number of the decisions matrix leaders face are complex rather than complicated. The right answer is not knowable until the organisation acts and observes what happens. Running a complex decision through a process designed for complicated ones produces false confidence: a decision that feels settled because the process was followed, when the real answer was never available to be found in advance.

Mistaking a polarity for a decision: Some of what gets brought to a decision making process is not actually a single choice to be made once and closed. It is an ongoing tension between two legitimate, competing needs, such as growth and efficiency, or local autonomy and global consistency. Treating a polarity as a decision produces a recognisable pattern: the same issue keeps reopening, no matter how decisively it was apparently resolved, because what was needed was an ongoing balance, not a verdict.

The practices that close the gap between decision rights and decision power

Map decision power alongside decision rights

A RAPID or RACI chart tells you who has the formal role in a decision. It does not tell you who, regardless of their formal role, has the practical ability to make the decision real or to quietly let it fail. These are often different lists, and the second one matters more in a matrix than the first.

The gap is not always about formal authority alone. Sometimes the right person occupies the right seat on the chart, gives genuine Agree or Input, and the decision still does not hold, because that person was never able to commit their whole function. They could speak for their own team. They could not speak for the wider division the decision depended on, and nobody checked whether that distinction mattered until implementation began.

A product director I worked with described a pricing decision where the RAPID chart named her marketing counterpart as Agree, and he gave it in the meeting. What had not been checked was that he led only the EMEA pricing team. The Americas and APAC teams reported elsewhere and first heard about the new pricing policy after it was finalised. His agreement had been genuine. It had also only ever covered a third of what the decision needed.

The operations director’s consolidation programme had a similar pattern. Two of the four divisional leads who had formally agreed to the decision reported through a separate line, with their own priorities and targets that had little to do with the programme. Even where they agreed personally, they could commit their own team but not the wider division the rollout needed. Their agreement in the room had cost them nothing. What happened afterwards depended on something the RAPID process had never asked: whether she had built enough trust with them, and with the people behind them, to make the decision matter once she had left the room.

Questions you can use to surface the gap between rights and power:

• Whose active cooperation does this decision depend on for execution, beyond the people formally named in the process?
• Of those people, who reports to the named Decider, and who does not?
• Can each person in the Agree or Input role genuinely commit their whole function, or only the part of it they personally lead?
• For anyone outside the Decider’s formal authority, what is currently making their cooperation likely, beyond goodwill?
• Who in this group has quietly let a previous decision die through inattention rather than open disagreement?

Practical actions:

• Before finalising the RAPID chart, ask each person in an Agree or Input role one direct question: can you commit your whole function to this, or only the part you personally lead? Treat any hesitation as the real answer, not the formal one.
• Build a decision power map alongside the rights chart, listing everyone whose active cooperation execution depends on and marking whether the Decider has any formal authority over them.
• For each name outside the Decider’s authority, agree a specific action to win their commitment directly, a conversation, a shared win, a standing update, before the decision is announced more broadly.
• Revisit the power map at the first major milestone, not just at sign off. Ask what the cost has been of any gap that turned out to be real, and who absorbed it.

Calibrate consultation to the level of risk, not the org chart

I have written elsewhere about the Waterline Principle, the idea, attributed to Bill Gore, that decisions risking a survivable, visible hole above the waterline can be made quickly and with real autonomy, while decisions risking a hole below the waterline, ones that could genuinely sink the wider effort, need broader consultation before they are made. In a matrix, this principle does work that the standard decision rights frameworks tend to skip.

A RAPID chart can correctly name a single Decider and still under consult, simply because the people who should have been asked were not formally required to be. There is also a deeper version of this problem worth naming: getting input from the people closest to the actual work, the ones who will feel a decision’s consequences first, matters as much as getting input from anyone with formal Agree rights. A clean RAPID chart can satisfy every formal requirement and still miss the perspective that would have surfaced the real risk.

A finance lead I coached saw this during a shared services restructure. The RAPID chart was followed, and the formal Agree group signed off without objection. Nobody had asked the regional finance teams who would process the new workflow day to day, because they did not hold a formal Agree role. Within a month, two regions had built workarounds, because the new process did not fit how their local regulatory reporting worked. The decision had been made correctly by every measure the chart could see. It had still missed the group whose daily reality it most affected.

The operations director’s programme had a version of the same gap. The formal sign off group never included the warehouse managers most affected by the new supplier model, because they did not have a seat on the RAPID chart for a decision at that level. Their early scepticism, voiced only to each other, turned out to be a fair prediction of where implementation would struggle.

Questions you can use to test whether a decision sits above or below the waterline:

• If this decision turns out to be wrong, who outside this group would feel the consequences, and how seriously?
• Could this decision be quietly corrected later without lasting damage, or would the damage be done the moment it was acted on?
• Have we consulted based on who is formally required to weigh in, or based on who would genuinely be affected if this goes badly?
• Have we heard from the people closest to the actual work, or only from those with a formal seat at the table?
• Are we treating this as routine because the process looked routine, rather than because the stakes actually are?

Practical actions:

• Before finalising any significant decision, ask the waterline question explicitly and write the answer down: above or below, and why. Do this separately from confirming who formally has the right to decide.
• For anything below the waterline, name at least one person closest to the actual work who has no formal role in the RAPID chart, and consult them directly before the decision is finalised.
• Where a decision sits below the waterline and the formal Agree group has signed off unanimously and quickly, treat that speed as worth a second look rather than a sign of confidence.
• After the decision lands, ask the people closest to the work one specific question: what did this miss that you could see coming? Use the answer to correct course before the gap becomes expensive.

Match the speed of the decision to whether it can be undone

A related but separate question is how reversible the decision actually is. Jeff Bezos’s distinction between one way doors, decisions that are difficult or costly to reverse, and two way doors, decisions that can be undone with relatively little cost if they turn out to be wrong, gives matrix leaders a second axis alongside the waterline question. The waterline asks how much damage a wrong decision would cause. Reversibility asks how much of that damage can be undone once it has happened.

The two questions work together. A decision that sits above the waterline and is also easily reversible can move fast, with light consultation, because even a mistake costs little and corrects easily. A decision that sits below the waterline and is also difficult to reverse deserves more consultation and more care, because there is no quick way back if it goes wrong. Many matrix leaders treat every decision as though it needed the same level of caution, which slows the genuinely reversible decisions down unnecessarily while sometimes rushing the ones that deserved more care.

A technology director told me about a multi country rollout where the team spent several weeks building consensus on which collaboration tool to pilot in one region, a decision that could have been reversed within days if it did not work. A separate decision, to retire a legacy data system entirely, was harder to reverse once executed and moved through approval in under a fortnight, because the group wanted to keep momentum. The two decisions had ended up the wrong way round. The pilot needed less time. The system retirement needed more.

The operations director’s own programme had a milder version of the same pattern. The choice of which supplier management software to trial in one division, easily reversed, went through three rounds of multi country sign off before anyone started using it. The decision to consolidate supplier contracts across the business, harder to unwind once signed, moved at the same pace as everything else in the programme, simply because that was the programme’s default rhythm rather than a deliberate choice.

Questions you can use to assess reversibility before deciding how to decide:

• If this decision turns out to be wrong, how easily and how quickly could we change course?
• What would it cost, in time, money, or trust, to reverse this decision once it has been acted on?
• Are we slowing this decision down because it genuinely needs care, or out of habit?
• Are we moving fast on this decision because it is genuinely reversible, or because moving fast simply feels more comfortable?

Practical actions:

• For every significant decision, classify it as a one way door or a two way door before deciding how much process it needs, and write the classification down alongside the waterline assessment.
• For two way door decisions, set a deliberate ceiling on consultation time, a number of days rather than a number of meetings, and hold to it.
• For one way door decisions, treat a unanimous, fast sign off as worth a second look rather than a reason to move faster. Ask whether the speed reflects genuine confidence or discomfort with slowing down.
• At a major milestone, review a few past decisions against this classification and ask honestly whether the pace matched the stakes. Use what you find to recalibrate.

Separate true decisions from ongoing polarities

Not everything brought into a decision making process is actually a single decision waiting to be made. Some of it is a polarity, an ongoing tension between two legitimate and competing needs that cannot be permanently resolved in favour of one side without losing something important from the other. Barry Johnson’s work on polarity management names this pattern: treating a polarity as a one time problem to be solved produces a predictable cycle, where the same issue keeps reopening no matter how decisively it appeared to be settled.

Common matrix polarities include growth against efficiency, local responsiveness against global consistency, and speed against thoroughness. A matrix leader who runs one of these through a standard RAPID process, naming a single Decider and expecting the matter to close, will often find the decision reopening within months, not because the process failed, but because the issue was never really a decision in the first place.

A regional marketing director described this cycle around how much campaign content should be set centrally versus adapted locally. Roughly every six months, the same RAPID process ran again. A Decider was named, and a decision was made, usually swinging towards either central control or local autonomy depending on who held the role that year. Within a couple of quarters, complaints resurfaced from whichever side had lost, and the process started again. It took someone pointing out that this had happened four times in three years for the group to recognise that what they kept calling a decision was actually a polarity that needed an ongoing rhythm of rebalancing, not a single ruling.

The operations director’s consolidation programme had its own version embedded inside it: the tension between standardising supplier processes across the business and allowing enough local flexibility for divisions facing different regulatory environments. Treated as a single decision, it produced the kind of unravelling described at the start of this article. Reframed as a polarity, with a recurring quarterly check in rather than a one time ruling, the tension stopped being a source of quiet resistance and became something the group managed openly together.

Questions you can use to tell a true decision from a polarity:

• If we make this decision, are we likely to be revisiting essentially the same question again within a few months, regardless of how well it was made?
• Is there a legitimate need on both sides of this issue that a single choice would lose entirely?
• Has this exact tension already been “resolved” more than once, only to resurface later?
• Would it be more accurate to describe this as something we need to keep rebalancing, rather than something we need to settle?

Practical actions:

• Before running a contested issue through RAPID, ask the group directly whether this has come up before in a different form, and if so, how many times. A pattern of repeated “resolutions” is a clear sign of a polarity in disguise.
• For anything identified as a genuine polarity, replace the single decision point with a standing review on a fixed cadence, and name it openly to the group as ongoing management rather than a decision still pending.
• Give each side of a recognised polarity an explicit owner whose job is to flag when the balance has tipped too far, rather than leaving the correction to whoever complains loudest.
• Resist the pressure to give a polarity false closure simply because the group wants the discomfort of an open question to end. Naming it honestly as ongoing, in writing, is more useful than a decision that will not hold.

Close the loop: From decision to delivery

A decision rights process has a natural endpoint. The Decider decides. The chart is filed. The meeting closes. In a hierarchy, that endpoint is close enough to execution that treating them as the same thing rarely causes serious problems. In a matrix, the distance between the decision being made and the decision being real can be considerable, and nothing in the formal process is designed to cross it.

The gap shows up in a particular way. A steering group signs off a decision, the minutes go out, and the leader assumes the work of deciding is done. What has actually happened is that the formal permission to act has been granted. Whether anyone acts on it, in the way that was intended, at the pace the decision assumed, depends entirely on things the RAPID chart was never designed to address.

A supply chain director I worked with described a regional restructure that had been signed off at board level, communicated to all affected functions in a detailed briefing document, and then largely not implemented for four months. The decision was never challenged. It was simply not prioritised by the people who needed to execute it, because none of them had been part of shaping it, none of them had made a specific commitment to a named milestone, and the leader who owned the decision assumed that board sign off and a briefing document were sufficient to make it happen. They were not.

What crosses the gap between decision and delivery is not governance. It is the four capabilities this series has explored from the beginning. Influence determines whether the people in the decision power gap actually choose to commit, rather than comply minimally or ignore the decision entirely. Networks determine whether the right people were genuinely reached before the decision landed, not just formally listed as consulted. Convening determines whether the decision was presented in a way that built ownership, or simply announced as a fact to be accepted. Accountability determines whether the commitments the decision created are specific, visible, and addressed directly when they start to slip.

There is also something quieter underneath all four: whether the people involved had any real history of working together before this decision arrived. A group with an established habit of collaborating across boundaries, with mutual respect and a willingness to compromise, tends to make decisions hold even when the formal process was imperfect. A group without that history can run a careful RAPID process and still watch the decision dissolve, because nothing about the relationship made the commitment feel real or the consequence of non-delivery feel shared.

The operations director’s consolidation programme recovered, not because the RAPID process was redone, but because she went back and did the work the process had skipped. She built direct relationships with the two divisional leads outside her formal authority. She convened a working session that let them shape part of the rollout rather than simply receive it. She made the resulting commitments specific and visible, and she addressed the first sign of slippage directly rather than letting it pass. The decision itself never changed. What changed was everything that happened after it, which is the part a decision rights framework was never designed to cover.

Questions you can use to close the gap between decision and delivery:

• Which of the four capabilities is most likely to be the weak point after this decision is made, and what will you do about it before launch rather than after?
• Who in the decision power map made a formal commitment but has not yet made a specific one, naming what they will do, by when, and what they need from others?
• Was this decision announced or convened? Did the people who need to deliver it have any hand in shaping how it lands in practice?
• What is the first sign that this decision is slipping, and who is responsible for naming it when it appears?

Practical actions:

• Before the decision is announced, run a delivery audit alongside the decision power map: for each person whose cooperation execution depends on, name the specific commitment you need from them, not their general support, and confirm they have made it explicitly rather than nodded along.
• Where the formal process ended with communication, go back and convene. Even a short working session that lets people shape one element of implementation, sequencing, rollout approach, local adaptation, changes the relationship to the decision from passive receipt to active ownership.
• Assign a named owner for the first ninety days of execution whose sole job is to track whether commitments are holding and surface slippage early. This is not a project manager role. It is a relationship role, and it needs someone with enough standing across the group to raise a slipping commitment directly rather than escalating upwards.
• At the first major milestone after the decision, ask the group one question: what would have made this easier to deliver from the moment the decision was made? Use the answer to design the next decision process better, not to assign blame for this one.

Wrapping up

Decision rights answer an important but narrow question: who is allowed to decide. They say nothing about whether the decision will survive contact with people who were never formally obligated to follow it, which is the situation most matrix leaders actually face. Frameworks like RAPID, the waterline question, reversibility, and polarity management each sharpen part of this picture, but none of them, on their own, closes the gap between the right to decide and the power to make a decision real.

That gap is closed the same way commitment is closed everywhere else in matrix leadership: through influence, networks, convening, and accountability. A decision rights process can tell you who holds the D. Only the four capabilities already explored in this series determine whether that decision becomes something more than a line on a chart.

Three questions for reflection

1. Think of a recent decision that was technically made but never really happened. Where exactly was the gap between who had the right to decide and who had the power to make it real?
2. Is there a recurring issue on your team that keeps getting “decided” and then reopening? Could it actually be a polarity rather than a decision?
3. For your next significant decision, who sits outside the formal process but holds real power over whether it succeeds, and what will you do differently to bring them in?

Inspiration

Bezos, J. (2016) Amazon shareholder letter. Seattle: Amazon.com.

Bhalla, V., Gandarilla, D. and Watkins, M. (2022) ‘How to make your matrix organization really work’, MIT Sloan Management Review, 64(1).

Johnson, B. (1996) Polarity management: Identifying and managing unsolvable problems. 2nd edn. Amherst, MA: HRD Press.

Mattessich, P.W., Murray-Close, M. and Monsey, B.R. (2001) Collaboration: What makes it work. 2nd edn. Saint Paul, MN: Amherst H. Wilder Foundation.

Rogers, P. and Blenko, M. (2006) ‘Who has the D? How clear decision roles enhance organizational performance’, Harvard Business Review, January 2006.

Snowden, D.J. and Boone, M.E. (2007) ‘A leader’s framework for decision making’, Harvard Business Review, November 2007.