The diagnosis nobody audited
Ask each function what is broken and every answer will be accurate. Add them up and the diagnosis is still wrong.
It happens every time a company decides something has to change. A transformation, a pivot, a new product line, a cost programme. The label changes but the first step does not, because somebody has to name what is broken before anything else can start. Get that wrong and everything after it is in the wrong order.
Nobody is lying, which is what makes it hard to catch. Each function describes what it can see from where it sits, and none of them looks further than what it owns. The problem gets defined at the depth the most senior person stopped digging, and everything downstream inherits it: the mandate, the hire, the programme, the budget.
Nothing in the operating model asks them to assemble a diagnosis between them. They report up, and somebody in the C-suite listens to partial accounts and decides what “this” is, in a room where nobody contradicts the person who decides their future.
By the time it reaches a job specification it has hardened into a sentence. “We need someone who can come in and fix this.” Buried in that line is a diagnosis nobody audited, and whoever is hired inherits it as their mandate.
Sometimes the named thing is the real one, and a person who is genuinely the problem is not a rare animal. One question separates the two and it almost never gets asked. If you replaced this person tomorrow, what would still be true on Monday?
Finance is often where it shows first. Everything every other function decides arrives in the ledger eventually, which makes the FP&A leader the easiest person in the building to point at.
So FP&A looks slow. Finance is running on four general ledgers from acquisitions nobody finished integrating, and no one has owned what the numbers are supposed to mean since the second one closed. Every close is four closes stitched together by people who know where the joins are.
Replace the person and the new one inherits four systems, the same calendar, and the same people holding the stitching. They spend their first year learning what the last person knew. Integration becomes another business case competing against work with revenue attached to it. Two years later the same conversation starts again.
Deloitte* asked 500 senior leaders in April, more than half of them at public companies above $5bn, and published the answer this September. 67% expected to hit their targets. 14% said they fully captured the value they were after on last year’s highest-priority initiative. Most of that gap gets written up as execution. Some of it was decided long before anyone executed anything.
I have spent more than fifteen years owning commercials in complex enterprises, much of it inside businesses built by acquisition. The multiple ledgers are never a secret. They surface in the architecture conversation every time, and every time the integration becomes a separate project, scheduled behind whatever is funded this year. Nobody carries it into the conversation where the FP&A leader is being discussed. Everyone who could has done the arithmetic on what raising it would cost them, and concluded it is not worth it.
It is not only finance. Go to market compresses margin on the hope that the next renewal pays for it. Delivery grants an exception because the relationship is developing. Procurement carries a tail nobody has time to consolidate. Every one is defensible by the person making it, which is how each one survives. The correction lands only after they compound, on whoever is standing closest. Replacing that person creates visible action and leaves the system that produced the result untouched.
A diagnosis set at the wrong depth does not get contested, it gets implemented. The person hired against it does exactly what the mandate asks for, and the organisation meets the cost of its own diagnosis for the first time. Nobody reads that cost as a diagnosis problem. It becomes a people problem, or a style problem, or suddenly the change is too large for the moment. Direct becomes too direct. “Hold my leaders accountable” acquires an exception for one particular leader. What was the need of the hour in January is too expensive and too big a change by September.
Most of how a company actually works is not written down anywhere. It lives in the people who built the workarounds, and over time the workaround becomes part of what their standing rests on. So when somebody arrives asking what is broken, they are asking the people who know the system best to expose the parts of it that exist only because they learned how to compensate for what does not work.
Nobody has to be hiding anything for this to fail. People defend what they built because it works, at least compared with what came before it. Defending it looks like expertise because much of the time it is expertise. That is also why transformation gets this wrong in two directions.
The pertinent example right now is the AI transformation. One version encodes the workaround, so something invented on a Tuesday to get round a broken handoff becomes permanent, faster, and much harder to unpick. The other redesigns from the documentation and never asks what the workaround was holding up, so the model stays clean until the first real exception breaks it. Both fail the same way. Nobody extracted the operating knowledge before deciding what to do with it.
This is how technical debt accumulates. Somebody builds around the workaround instead of fixing the process underneath, because building around it is cheaper this quarter and the real fix sits in somebody else’s budget. Do that four or five times and the workaround is holding up the building.
The more expensive version comes from above. Resistance in the ranks has a shape you can name and take to the person above it. A sponsor whose attention has moved has no shape at all. The C-suite can still believe in the programme while no longer behaving as though it comes first. Nothing is ever revoked. “Now is not the time” appears in a meeting nobody logged, and a mandate that was never withdrawn quietly stops being one. The agreement in that first room only ever held while the person at the head of the table was sitting in it. A mandate is worth what it costs the person who issued it, and most are never tested until the change takes something from a business, a function, or a person the sponsor is unwilling to disrupt.
Convenient and necessary are very different things, and most organisations protect the first while calling it the second. What you are actually willing to do shows up as an order of operations. In a business built by acquisition the dependencies run like this. Finish the integration before you touch the ledger, because until the ledger is one thing you are reading four versions of the same problem. Settle the entity structure before anybody moves, because a person moved ahead of the structure meant to hold them triggers cost in a jurisdiction nobody modelled, and the expense line blows out on a decision every function made correctly. The ERP is the one everybody reaches for first and it belongs close to last, because a system built on an unanswered question only makes the question faster.
Nobody can see the whole order from the starting line, and the first job is finding it rather than declaring the path to victory. The stages move and overlap, and the right order in one business is the wrong one in the next. What does not move is that you cannot choose which to skip, because each is holding something up for the one behind it. Rush one and the cost never lands there. It appears three stages later, in somebody else’s budget, as a surprise.
Done in order, the spend is heavy at the front and there is nothing to announce for a while. That is the only version of the plan with no early win in it, so it is the first thing to go when attention moves. Sequencing loses to visible action, and replacing a person shows up this quarter.
There is a test for whether a room ever committed and it is not a survey. Everybody says they disagree and commit, and I have run teams that way for years because decisions only matter when they survive discomfort. The tell is operational: calls start getting missed and calendars fill, the owner sends a delegate, and something breaks in the handoff between two functions with neither one quite responsible. The strategy has not changed, only what anyone is prepared to spend on it.
So ask what order the fix has to happen in. Ask the person issuing the mandate, and ask before anyone is hired to deliver it. If the answer is a set of workstreams running in parallel with no dependency named between them, the work has not been done. Whoever takes the job finds out which stage was holding up which, in the quarter it becomes their fault.
* Deloitte, 2026 Enterprise Cost Transformation Survey. 500 C-suite, vice president and other senior business leaders, surveyed April 2026; more than half at publicly traded organisations with annual revenue above $5bn. Published 14 September 2026.