You Delegated the Task. You Kept the Decision.

You handed the supplier renegotiation to your operations lead a month ago. She has done the work. The comparison is built, two quotes are shortlisted, the incumbent has been told a review is under way. On Thursday it arrives back in your inbox under a covering note of one line: both options look workable, let me know which way you want to go.

So it is back. Not the labour — she did all of that, and did it better than you would have. Only the part where somebody has to choose.

This is the most common failure in delegation and it almost never registers as one, because nothing visibly went wrong. No deadline slipped. Nobody needed chasing. The work came back finished. And the thing you were trying to get off your desk is on it again, in the only form that was ever expensive.

Why do decisions keep coming back to me after I've delegated?

Because you delegated the work and kept the choice, and the choice was the only part consuming you.

Doing a task costs hours. Owning a decision costs something else: the attention it holds while it is open, the reserve it draws down when it closes, and the exposure of having been the person who called it. Hand over the hours and keep the ownership, and you have moved the cheap component and retained the expensive one. Your calendar clears slightly. Your head does not clear at all.

William Oncken and Donald Wass described the mechanism in the Harvard Business Review in November 1974, in an article that became one of the most reprinted the magazine has ever run. They called the open item a monkey, and they tracked which back it sat on. A member of staff brings a problem. The manager says: let me think about it and get back to you. In that sentence the next move transfers. The person who raised it walks away unencumbered and periodically checks progress; the manager now owes them an answer. Oncken and Wass named the resulting load subordinate-imposed time, and observed that managers acquire it voluntarily, one reasonable sentence at a time.

Fifty years on the sentence has changed sides. It is no longer the manager saying let me get back to you. It is the covering note saying let me know which way you want to go. The direction of transfer is identical. The task was done at their desk and the decision was returned to yours, and both of you experienced that as good work being handed up for sign-off.

Nothing about this is a discipline problem, on either side. She was not being timid. She had no authority to conclude the matter, so concluding it was not among the things she could do. She did everything she could do, and then she did the only remaining thing available to her, which was to ask.

What is the difference between delegating a task and delegating a decision?

A task is who does the work. A decision right is who is permitted to commit the organisation to an answer without coming back for approval. They are separate grants, and only one of them relieves you.

Delegation without decision rights is not delegation. It is task assignment with a compulsory return leg.

The distinction is not a piece of consultancy vocabulary; it is the most heavily evidenced finding in the literature on organisational execution. Gary Neilson, Karla Martin and Elizabeth Powers published an analysis in the Harvard Business Review in June 2008 drawing on survey responses from more than 125,000 employees across roughly 1,000 organisations in over 50 countries. They were looking for what separates companies that execute their strategy from companies that announce it. The answer was not incentives and it was not the organisation chart. The single most common attribute of the strongest executors was that people were clear about which decisions and actions they owned. Clarifying decision rights and improving information flow did more than restructuring, and restructuring is what most firms reach for first.

Bain's research points the same way from the financial end. Across a survey of nearly 800 companies, the firm reported a 95 per cent correlation between organisations that excel at making and executing key decisions and those in the top tier of financial results, measured by revenue growth, return on capital and total shareholder return. Marcia Blenko, Michael Mankins and Paul Rogers built Decide & Deliver on that finding. Their working unit is not the strategy or the structure. It is the individual decision, and specifically who holds it.

So the question worth asking about any delegated piece of work is not how it is progressing. It is: if this person reaches a conclusion I would not have reached, does it stand? If the answer is no, or if the answer is that it depends on what the conclusion turns out to be, then the decision never moved. It was on loan.

Why doesn't hiring more senior people fix it?

Because seniority is a description of someone's judgement, and what returns work to your desk is the absence of a stated right, not the absence of capability.

A very experienced hire arriving into an unstated arrangement will do what any careful professional does: ask. They have no map of where their authority ends, and the cost of overstepping in a new post is high while the cost of checking is nil. Ambiguity always resolves upwards, because upwards is the safe direction. Six months in, the resolution has hardened into a habit that neither of you named, and you conclude that the hire has not stepped up. What actually happened is that nobody ever said where they were permitted to stop.

This is also where the formal machinery of most companies runs out. McKinsey's organisation practice sorts consequential decisions into three kinds: big-bet decisions, which are rare and shape the company's future; cross-cutting decisions, which recur and get made in cross-functional forums; and delegated decisions, which are frequent, individually low-risk, and best handled by the person closest to the work. Businesses tend to build careful process around the first two and nothing at all around the third. Delegated decisions fall through the gap, and so they default back to whoever has traditionally decided everything.

The volume is the point. McKinsey's 2019 survey of more than 1,200 managers found they spend around 37 per cent of their working time making decisions, with more than half of that time judged to be spent ineffectively. That waste is not sitting in the acquisition or the restructure. It is in the vast body of small, recurring calls that keep arriving at a desk they should never have reached — each one individually trivial, each one taking a slice of the same finite reserve, none of them big enough to be worth an argument. It is the same trap as permanent firefighting: the load is made of items too small to refuse, and it is crushing precisely because no single item deserves attention.

What are decision rights, and how do you actually hand one over?

A decision right, handed over properly, is four things written down and none of them is a conversation about trust.

The class of decision, named specifically enough that both of you can tell whether a given item is inside it: supplier terms under a stated value, refunds, hiring below a certain grade, discount within a range. The person, singular. Rogers and Blenko's finding, from the RAPID work they set out in the Harvard Business Review in January 2006, is unambiguous on that count: one decider, and a deliberately short list of anyone holding a veto, because a decision with two owners is a decision with none. The limits, stated as thresholds rather than as judgement — above this figure it comes to me, below it, it does not. And the reporting line afterwards, which is that you are informed, not consulted.

Then the part that decides whether any of it holds. You have to state, in advance, what happens the first time they decide something you would not have decided. Because they will, and it will not be a disaster, and it will be visibly not what you would have done. If the answer is that you quietly take the decision back — no announcement, just a returned email and a suggestion — the grant is over, and everyone can see it is over, and the next twenty items in that class will be routed to you for safety. A decision right you can revoke case by case, silently, was never a right. It was a rehearsal.

This is why the standing objection to it is worth naming plainly. Most people do not withhold decision rights because they think their staff are incapable. They withhold them because they cannot see the consequences quickly enough to sleep, and the veto is what they hold instead of visibility. The fix for that is not to keep the veto. It is to build the visibility, so that the consequence of a delegated call arrives in front of you on a schedule you set rather than being something you have to go and check. Operational Exposure is that single mechanism: a recurring, non-dismissible view of the number you would otherwise only look at when worried. Once the number comes to you on its own, you no longer need the approval step to feel safe, and the approval step is the whole return leg.

The wider job — mapping which classes of decision are actually reaching you, setting the thresholds, and writing the escalation rule so that exceptions run on a stated trigger rather than on someone's nerve — is what Operational Protocol is built for. Five levers, each aimed at a specific way authority silently returns to the top. None of them require you to become more relaxed about being wrong, which is fortunate, because you will not.

What changes when the decision moves?

The email stops arriving.

That is the whole of it, and it is worth being unromantic about the scale. You do not get a transformed organisation. You get one class of item that no longer requires you, then another, then a third, and the compounding happens quietly across months. The reserve you were spending on other people's choices goes back into the small number of decisions nobody else can make, which is the only thing that was ever your actual job.

The supplier note is the test case. You could answer it today. It would take four minutes and it would be the right call, and in six weeks something structurally identical would land under a different subject line. Or you could answer it once, differently: name the range within which she chooses and tells you afterwards, name the figure above which it comes back, and say out loud what you will do when she picks the option you would not have picked.

The second reply takes longer and ends a category. The first one takes four minutes and buys you six weeks.