When a program stalls, the instinct in most conference rooms is to find out what everyone agrees on. Get leadership, middle management, and the people doing the work into alignment, the thinking goes, and the plan will follow. That instinct is backwards. Consensus is not where the diagnosis lives. The gap is.

Lay three accounts of the same stalled program side by side, leadership's, middle management's, and the account of the people closest to the work, and they rarely describe the same problem. Leadership will point to a team that seems slow or disengaged. Middle management will describe resourcing and timeline pressure. The people actually doing the work will usually name something narrower and more specific: a decision that never gets made, an approval that sits unresolved, an owner who was never actually assigned. These are not three versions of the same story. They are three different stories, and the distance between them is the finding.

The natural response to that distance is to average it out, to look for the version of events that most people can agree to and build a plan around that. This is exactly the wrong move, and it is the move almost every internal process defaults to. A consensus account is, by construction, the account that offends no one, which means it has already been sanded down to remove whatever was uncomfortable enough to disagree about. The widest gap in the spectrum, not the narrowest, is where the real system failure is hiding, because that gap is the one place someone's account is actually being suppressed or overridden by the layer above them.

This is why a listening tour has to start with leadership and then go past it, rather than stopping there. Leadership's account is not the brief. It is one data point among several, and often the least reliable one about what is actually happening at the point of execution, because leadership sits furthest from it. The people closest to the work know what is broken. They are usually heads-down delivering their current commitments, inside their own silo, with no mandate and no occasion to step back and describe the system as a whole to anyone who is listening. Someone has to go get that account directly, not filtered through a management layer that has its own incentive to smooth it over.

Middle management's account deserves particular scrutiny here, not because the people in it are dishonest, but because their position in the org chart makes their account structurally the least reliable of the three. It has to protect what leadership believes above it and explain what the team is actually experiencing below it, and an account built to do both of those things at once is rarely built to be accurate about either. Treating it as the tiebreaker between the other two accounts, rather than as its own distinct and compromised data point, is one of the most common ways a real diagnosis gets lost before it is ever made.

Once the gap is mapped, the fix has to be built at the gap, not around it. This is the part that separates a diagnosis from an intervention. It is not enough to know that leadership believes one thing and the team experiences another. The system has to be rebuilt at the exact point where those accounts diverge, and it has to be rebuilt with the people closest to the work, not delivered to them as a finished document. A process redesigned without the people who will run it is a process that drifts back toward the old failure, because nobody who has to live inside it had a hand in building it.

None of this requires more oversight, more reporting, or a reorg imposed from the top. It requires going to look at the actual gap instead of averaging past it. Most operational dysfunction gets diagnosed as a people problem because the people making the diagnosis stopped at the account that was easiest to collect, usually leadership's own. The account that is hardest to collect, from the people who are actually inside the work, is almost always the one that names the real problem. Start there, and the fix that follows tends to hold. Start with consensus, and the same stall reappears in a different disguise.