- The Question, Restated: Where Are Decisions Waiting?
- Queue One: Before Anyone Proposes Anything
- Queue Two: Information That Would Not Change the Answer
- Queue Three: Waiting for the Meeting
- Queue Four: An Authority Nobody Has Named
- Queue Five: Decided but Not Yet Real
- How Do You Tell Which Queue You Have?
- What to Change First
- What Not to Do
- Frequently Asked Questions
- About the Stagnation Assassin
The question people ask is how to make decisions faster. That is the wrong question, and asking it produces the wrong interventions: shorter meetings, tighter agendas, a decision-making training course, an exhortation from the top about bias for action. All of those address the time spent deciding, which in a large organization is almost never where the time goes.
The better question is where decisions are waiting. Take any decision that took eleven weeks and break the eleven weeks into its parts. The time actually spent thinking about it, discussing it, and analyzing it will usually total a few days. The rest is queue. And queue is a structural property of an organization, not a behavioral one, which is why exhortation does nothing and why the same company can be genuinely urgent in its culture and genuinely slow in its throughput.
There are five places decisions queue in a large organization. Each has a different detection signature and a different fix, and applying the fix for one to another is the most common reason decision-speed initiatives fail. What follows is how to tell them apart.
The Question, Restated: Where Are Decisions Waiting?
Decision latency is elapsed time, not work time. The gap between the two is the queue, and in most large organizations the queue is the overwhelming majority of the total. Measuring the work time and trying to compress it is optimizing the small number while ignoring the large one.
Elapsed time versus touch time
Manufacturing has understood this distinction for decades. A part that takes four minutes of machining and six weeks to get through the plant does not have a machining problem. Nobody would respond to that by trying to make the cut faster. They would look at where the part sits between operations, and they would find queues, batching, and waiting for a resource that was busy with something else.
Decisions behave identically and get analyzed completely differently. The instinct with decisions is to make the cut faster, which is why so many organizations run decision-making workshops and emerge with the same elapsed times.
Why “faster” is the wrong target
Setting a target of faster decisions without naming the queue produces one predictable outcome: people stop bringing decisions forward, because the fastest way to have a short decision cycle is to arrive with the decision effectively already made. The metric improves and the organization gets slower in the part nobody is measuring.
So the target is not speed. The target is identifying which of the five queues holds your decisions, and then removing it. Speed is the result.
Queue One: Before Anyone Proposes Anything
The invisible queue. A problem is visible to somebody for months before it reaches an agenda, because raising it early carries a cost and raising it late does not. This queue does not appear in any process measurement, since nothing is formally in progress yet.
How to detect it
Take six recent decisions and ask the people closest to the issue when they first knew. Not when it was escalated. When they first knew. If that date is consistently months before the agenda date, this is your dominant queue, and nothing you do downstream will matter much.
A second signature: proposals that arrive fully formed, pre-defended, with a single recommended option and the uncertainty already removed. That is not thoroughness. That is a rational response to an environment where bringing forward an open question means months of being questioned about it.
The intervention
Change what happens to the person who raises something early and uncertain. Concretely: a standing slot where items can be raised without a recommendation, an explicit rule that the raiser is not the owner unless they volunteer, and a visible instance of a senior person raising something half-formed and being treated well for it.
The second half of the intervention is subtler and matters more. Separate the act of raising an issue from the act of owning it. In most organizations those are fused, so raising something means acquiring it, and acquiring something you do not have the resource to fix is a rational thing to avoid. Break the fusion and the raising rate goes up within a quarter.
This queue takes a year to fix and there is no process substitute. It is the most expensive of the five and the least addressable by structure, which is also why it is the one that survives reorganizations, new leadership, and every process improvement programme that arrives in between.
Queue Two: Information That Would Not Change the Answer
The analysis queue. Work is commissioned, and it is real work done seriously, but the result would not alter the decision regardless of what it showed. This is the most socially defensible delay available, which is exactly why it is the most common.
How to detect it
One question, asked before any analysis is commissioned: what result would change our decision? If nobody can name a threshold, the analysis is not decision support. It is reassurance, and reassurance has a cost measured in weeks.
The other signature is the analysis that arrives and confirms what everyone expected, followed by a decision identical to the one that would have been made before it started. Look back at the last five commissioned analyses and count how many changed an outcome. The number is usually one.
The intervention
Require a stated decision threshold with every analysis request. One line: if the answer is above X we do A, below X we do B. Requests that cannot produce that line get answered from judgment instead, immediately, and the analysis is either dropped or run in parallel with the decision rather than ahead of it.
Running analysis in parallel rather than in series is the underused half of this. Where the work is genuinely worth doing but the decision cannot wait for it, make the decision now with a scheduled review at the point the analysis lands. You get the decision and the learning, in the order that costs least, and the analyst’s work stops being a gate and becomes what it should have been, which is a check.
Watch for the version of this queue that hides inside a legitimate request. Someone asks a fair question in a review, the answer is not to hand, and an action lands on the analyst without anyone deciding that the decision now waits for it. Nobody chose the delay. It attached itself to the item during a forty-second exchange, and it will cost three weeks.
Queue Three: Waiting for the Meeting
The cadence queue. The decision is ready, the authority exists, and the forum that makes it meets monthly. Average wait is half the cycle, and if the item misses the agenda because the meeting ran long, it is another full cycle.
How to detect it
Plot the decision dates for the last quarter. If they cluster tightly around the same days of the month, your latency is set by your calendar rather than by your work, and the fix is structural and cheap.
The second signature is agenda overflow. If items are routinely deferred to the next session, the forum is capacity-constrained, and every deferral adds a full cycle to something that was already ready.
The intervention
Two moves. Increase the frequency of the forum that holds the most-deferred class of decision, which usually means a shorter weekly slot rather than a longer monthly one. And create an out-of-cycle path with a named threshold, so a genuinely ready decision that would otherwise wait three weeks can be made in writing between sessions by the same people.
The out-of-cycle path needs a stated threshold or it will not get used. Give people a rule they can apply without asking permission to apply it: above this dollar value, or affecting this class of customer, a decision can be made in writing between sessions by the named decider plus one consulted party. Without the threshold, using the path requires a judgment about whether it is important enough, and most people will not make that judgment on their own behalf.
This is the cheapest of the five to fix and it is where I would start in almost any organization, not because it is the largest queue but because fixing it is fast and visible, which buys the credibility needed for the harder ones. The structure it sits inside is the operating cadence.
Queue Four: An Authority Nobody Has Named
The rights queue. The decision is ready and nobody is certain who makes it, so it circulates. Each person it reaches has a legitimate reason to involve someone else, and the circulation is polite, reasonable, and indefinite.
How to detect it
Ask three senior people who decides a specific class of decision. If you get three answers, or three versions of “it depends,” you have found this queue. The follow-up that confirms it: ask who decided the last one of that class, and watch the pause.
This queue is strongly associated with cross-functional decisions, where the authority genuinely is distributed and nobody has ever written down how it resolves. Harvard Business Review’s work on cross-silo leadership is useful on why lateral coordination degrades in exactly these places, though the operational fix is narrower than the cultural framing suggests.
The intervention
Write down the decision rights for the twenty decisions that recur. Who decides, who must be consulted, who is informed after. Twenty is enough; attempting a comprehensive map produces a document nobody reads.
The rule that matters more than the map: when authority moves down, the information and the context have to move with it, or the person now holding the decision will escalate it back up out of reasonable caution, and you will have added a step rather than removed one. The full treatment of that assignment problem is in who should actually make this decision.
One detail that determines whether the map works at all: state the limits. Authority granted without a boundary is read narrowly, because the safe interpretation of an unbounded grant is a small one. A dollar figure, a customer tier, a time horizon. Three of your twenty decisions will need a number attached, and those three are where most of the value sits.
Queue Five: Decided but Not Yet Real
The implementation queue. The decision was made weeks ago and nothing has changed, because it was recorded as agreement rather than as an assignment with a date. Everyone in the room believes it is done. Nobody outside the room has been told in a form they can act on, and no single person is late.
How to detect it
Compare the date a decision was made against the date behavior changed. If that gap is routinely longer than the deciding took, the queue is here. The signature phrase is a leadership team that believes a decision is done and an organization that has not yet heard about it in actionable form.
A related signature: the same decision appearing on an agenda twice, months apart, with nobody quite sure whether it was settled the first time. That is a record-keeping failure presenting as indecision.
The intervention
Every decision leaves the room with three things written down within the hour: what was decided, who owns the first change, and the date that change happens. Two lines, published to everyone affected. This is the single cheapest intervention across all five queues and it eliminates an entire category of apparent slowness.
Resist the instinct to build a system for this. A shared document that the person who chaired the meeting updates before leaving the room outperforms any tool, because the tool introduces a step between the decision and the record, and that step is where the practice dies. The discipline is temporal rather than technological: within the hour, or it does not happen.
How Do You Tell Which Queue You Have?
Take six recent decisions, mark four dates on each, and the queue identifies itself. First-known date, agenda date, decision date, behavior-change date. Whichever gap is largest names your dominant queue, and the intervention follows from the name.
Do the exercise on decisions that were actually made rather than on ones still open, or the last date is unavailable and the whole grid collapses to guesswork. Six closed decisions taken as they come, not selected, is enough. Selecting them defeats the point, and the temptation to select is real, because everyone remembers the fast ones more fondly.
The reason this matters is that the queues mask each other. Fix the meeting cadence while the real queue is upstream, and the decisions simply arrive later and still get made on the new faster rhythm. The total elapsed time is unchanged and the initiative reports success.
What to Change First
Start with queue five, because it is nearly free and it takes a week. Then queue three, because it is structural and visible. Then whichever of one, two, or four your dates actually pointed to.
The sequencing logic is about credibility rather than about size. Queues one and two are the largest in most large organizations and both take quarters to shift, and an organization that has just watched two cheap fixes work will extend more patience to a slow one. An organization that has watched a twelve-month behavioral programme produce nothing measurable will not.
There is one exception to that order. If asking three senior people who decides produces three different answers, fix that first regardless of what the dates say, because unclear authority contaminates the measurement of every other queue. A decision with no owner will present as an analysis problem, or as a scheduling problem, or as nothing at all, and you will spend a quarter fixing the wrong thing.
Set the measurement up before you change anything. The four dates on six decisions, repeated quarterly, is the whole measurement system. It takes two hours and it is considerably more honest than any survey about whether people feel the organization is decisive.
What Not to Do
Three interventions to avoid, all of which are commonly attempted and all of which reliably make the measured number better and the real elapsed time worse. They fail for the same underlying reason: each one acts on the visible part of the process rather than on the queue.
Do not announce a decision-speed initiative. Naming it guarantees that the visible behavior changes and the queue does not, because the easiest way to comply is to compress the part that was already short. Worse, it teaches people that decisions are now being timed, which increases the incentive to arrive pre-decided and pushes the wait upstream into the one queue you cannot see.
Do not delegate more decisions without also moving the information. This is the single most common failed intervention: authority is pushed down, the people receiving it lack the context to be confident, and they escalate back up with an added step of politeness. Latency increases and the delegation gets blamed on the people rather than on the design.
Do not run the fix as a project with a steering group. A queue-removal change is small, structural, and boring: a shorter forum, a written threshold, a two-line record. Wrapping it in project governance adds a decision layer to an initiative whose entire purpose is removing decision layers, and the irony is lost on nobody.
Do not attempt a comprehensive decision-rights map. The twenty recurring decisions are worth writing down and the other four hundred are not. Comprehensive maps take six months to produce, are obsolete on delivery, and consume exactly the credibility the cheap fixes were supposed to build.
Frequently Asked Questions
Is decision latency a culture problem or a structure problem? Mostly structure, with one exception. Queues two through five are structural and fixable with design changes inside a quarter. Queue one, the waiting before anything is proposed, is behavioral and takes about a year, because it is driven by what happens to people who raise uncertain things early.
How do you measure decision latency? Take six recent decisions and record four dates for each: when someone first knew about the issue, when it reached an agenda, when it was decided, and when behavior actually changed. The largest gap names the dominant queue. Repeat quarterly.
Why does delegating more decisions often make things slower? Because authority moved without the information and context moving with it. The person now holding the decision escalates it back up out of reasonable caution, which adds a step rather than removing one.
Which decision queue should you fix first? The implementation queue, because writing down what was decided, who owns the first change, and when it happens costs almost nothing and works within a week. Then the cadence queue. Then whichever of the remaining three your own dates identified.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated $3B in shareholder value across Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, where he serves as Global VP of Product Strategy and Marketing. Known as The Stagnation Assassin, he is the author of two published books: The Unfair Advantage: Weaponizing the Hypomanic Toolbox and Stagnation Assassin: The Anti-Consultant Manifesto. His blog is published in 15+ languages and read by operators worldwide. Bring him to your stage via his speaking page or connect with him on LinkedIn.
Decision Velocity Audit. Run the four dates on six decisions and send me the grid. I will tell you which queue is dominant, which interventions are wasted on it, and what to change first. Free, and there is no pitch at the end of it. Request the audit here.