A managing director of a 20-person services firm at her Tuesday operations review. The team has prepared three recommendations. She approves two. On the third she asks for a revised analysis. The team knows what this means. The decision is now indefinitely deferred. It will be discussed again next Tuesday and will probably get made when she has a flight to catch and decides quickly to clear the inbox.
She has done this fortnightly for two years. She thinks she has delegated. She has not. She is asking herself why nobody on the leadership team grows into real ownership.
Why does the matrix fail even when it is well-designed?
The matrix is fine. The thresholds are explicit. The four tiers are clearly defined. The team has read it. The behaviour is the problem. The team makes a recommendation, the founder asks for more information or a revised analysis, the team reanalyses, the founder says “I am not sure this is right,“ the decision is deferred, and the team returns to waiting for founder input. This is abdication disguised as involvement.
The founder feels rigorous. They are doing due diligence. The team feels something different. They feel their judgement being questioned, repeatedly, on decisions the matrix says they own. After two or three rounds of this they stop bringing decisions and start bringing problems for the founder to solve. The matrix becomes documentation of a system that is not running.
The hidden message in every revised-analysis request
To the founder, asking for more information feels like rigour. To the team, it reads as “I do not trust your judgement on this.“ After two or three iterations of this, the team learns the safer move is to bring the founder a problem, not a decision. They stop pre-deciding. They stop offering recommendations with conviction. They wait.
The behaviour the founder reads as the team not stepping up is the rational response to the signal the founder has been sending. The first time this happens it might be defensible. The recommendation might genuinely have lacked information. The third time it happens, on three different decisions in a row, the message is no longer about rigour. It is about who actually has authority. The team reads the second message clearly even when the founder does not.
Permission to decide, stated explicitly
The shift sounds like this. “You are authorised to make this decision. I need to know the outcome and the reasoning. I will not reverse the decision after the fact unless you have violated an explicit boundary condition, or the outcome has created material harm.“ This text matters. It is the contract.
Without it, the team has to guess where the line is, and they tend to guess on the side of asking the founder again. Once the team holds this paragraph as a literal commitment, behaviour changes within weeks. The recommendations become decisions. The decisions get made. The reasoning is captured. The founder is informed, not consulted. If the founder cannot say this paragraph and mean it, the four-tier matrix remains aspirational rather than operational.
The boundary conditions, named
Three boundaries hold the permission. The decision is reversible (a two-way door, not a one-way door). It does not violate a legal or ethical constraint. It does not impact more than one functional area without cross-functional consultation. Anything inside these stays decided. Anything outside escalates. The boundary list is short on purpose. Long lists of conditions become a different version of “ask me first.“
The boundaries are written down and reviewed quarterly. The team can read them in fifteen seconds. They are the framework inside which authority lives. Any time the founder feels the urge to intervene, the test is “did the team breach a boundary.“ If yes, the intervention is legitimate and named as a boundary breach. If no, the intervention breaks the matrix.
The single reversal that breaks the system
If the founder reverses a tier three decision once because they did not like the conclusion (not because a boundary was breached), the team learns the matrix is provisional. The next decision they could have made, they recommend. The decision after that, they ask. The dependency returns inside two months.
The reversal does not need to be loud or confrontational. A polite “I have thought about this and I would prefer we go a different way“ is enough. The team reads it. Founders find this rule uncomfortable because it asks them to live with decisions they think are suboptimal. The trade-off is real. Allowing one wrong decision to stand is the cost of the team owning the next ten. Reversing the wrong decision saves one outcome and breaks the next ten. The math heavily favours not reversing, and most founders need a few cycles to feel that math intuitively.
The substitute behaviour
When the founder disagrees with how a decision was made, the feedback refines the decision process for next time. It does not reverse the current decision. The format is a debrief, not an intervention. “Here is what I would have considered. Here is the framework I would have applied. For decisions of this shape going forward, please walk through these steps.“ The decision still stands. The next decision benefits from the framework.
This feels passive when the decision is wrong. It is the behaviour that builds the team’s authority. Founders who hold to the substitute behaviour for three to six months find that the team’s decision quality improves measurably and the founder is asked to intervene less, not more. The discomfort in the early weeks is the cost of building the layer.
When a decision genuinely needs reversing
Sometimes a decision needs reversing because the outcome has created real harm or the team breached a boundary unknowingly. When this happens, name it as a boundary breach explicitly. The framing matters: “I am reversing this because the contract change you authorised crossed condition X. The fix is to add the cross-functional check to the standard process.“
This framing preserves the matrix. The reversal is legitimate because a boundary was breached, and the system improves because the breach is now caught earlier. The matrix says cross-functional review is required for changes of this size, and the team missed that.
Compare with the reversal framed as “I have reconsidered and I think we should do something different.“ The first framing is matrix-preserving. The second is matrix-destroying. Same outcome, different message. The team reads the difference.
Track yourself for two weeks
Keep a spreadsheet for ten working days. Every time you reverse a tier three or tier four decision, every time you ask for a revised analysis on a tier three recommendation, every time you push a decision back to the team for reconsideration, write it down. The category, the date, the boundary breach (if any), and the actual reason.
At the end of two weeks read the column titled “actual reason.“ The honest answers usually fall into three buckets: a real boundary breach (legitimate reversal), a quality concern about how the decision was reasoned (legitimate feedback for next time, illegitimate reversal of this one), or simple disagreement with the team’s call (illegitimate reversal). The proportion in each bucket is the diagnosis. If the third bucket is filling up, the matrix is theatre and the team already knows it.
If you would like a second pair of eyes on what the audit is telling you, book a conversation.
