A pricing decision sits in review for eleven weeks. A product sunset decision runs through four steering committees over a quarter. A geographic exit is debated across six monthly leadership meetings before anyone commits. In each case, the decision is eventually made, and in each case, the executives involved would describe the process as careful, thorough, and aligned. Nobody in the room would characterize what happened as slow. And yet the window in which the decision would have produced its maximum value has closed, often without anyone noticing.
This is the essential quality of decision latency as a management problem. It does not show up as a line item, it does not produce a visible failure, and it is almost never measured. A slow decision that is eventually made looks, in retrospect, like a good decision. The counterfactual (what would have happened if the same decision had been made three months earlier) does not exist in the record. The executive team has no way to see what the delay cost them, and so they do not know to value speed.
The Bain research on decision effectiveness has been remarkably consistent over a decade. Companies in the top quartile for decision quality, speed, and execution generate total shareholder returns that meaningfully outpace those in the bottom quartile. The spread is large. What is striking is how few companies treat decision performance as a measurable, manageable attribute of the business in the way they treat sales performance or operating margin.
- Bain's decision effectiveness research finds that companies strong on decision performance outperform peers by roughly 6 percentage points of annual total shareholder return.
- McKinsey reports that only 20 percent of executives believe their organization is good at making decisions quickly.
- Senior executives spend an average of 37 percent of their time on decisions, and more than half of that time is described as ineffective.
The relationship in Figure 1 is not a controlled experiment and does not establish causation. Fast-decision companies may simply be better-run companies on many dimensions. But the pattern replicates across studies, industries, and geographies, and the effect size is large enough that treating it as noise requires an unusually high burden of evidence.
Why consensus feels like the safer choice
The executive instinct to pursue consensus is rational at the level of any individual decision. Consensus reduces blame. A decision taken by the leadership team, with the inputs of every relevant function, with documented alignment, is a decision that no single person owns and therefore no single person gets fired for. The political economy of the executive suite rewards this configuration.
It also feels like good governance. Pulling in more voices looks like diligence. Waiting for the data to settle looks like rigor. Holding another meeting looks like care. Each of these has a legitimate purpose some of the time. What the discipline of decision effectiveness reveals is that most of the time, they are substitutes for making a decision, not preconditions for making a good one.
The problem is that the cost of consensus is carried by the company, not by the individuals who produce it. When a decision sits for ninety days, the product team has been building against the wrong spec for ninety days, the sales team has been selling against the wrong narrative for ninety days, the finance team has been forecasting against an assumption that no longer holds. Those are real dollars. They do not appear in any meeting because no line item exists for the cost of deferred decisions.
The chart above does the most useful work of any diagnostic on this problem. The analytical phases (data gathering, modeling) consume roughly a quarter of the elapsed time. The alignment and review phases consume more than half. The decisions are not slow because the analysis is hard. They are slow because the alignment is hard, and the alignment is hard because the organization has substituted consensus-seeking for accountability.
A slow decision that is eventually made looks, in retrospect, like a good decision. The counterfactual does not exist in the record.
Alignment is not agreement
The single most useful distinction in decision architecture is between alignment and agreement. Alignment means the organization will execute a decision even if some of its participants would have made a different one. Agreement means everyone in the room endorses the decision on the merits. These are different objectives, and conflating them is the primary source of decision latency in large organizations.
Companies that pursue agreement will keep iterating until every voice in the room nods yes. Companies that pursue alignment will debate vigorously, commit explicitly, and execute as a unit. Bezos's "disagree and commit" is not a catchphrase. It is a decision protocol that distinguishes the two objectives and tells the organization which one it is optimizing for.
In practice, the test is simple: after the decision is made, can any executive in the room ever say, in any forum, "I didn't really agree with that"? If the answer is yes, the company was pursuing agreement and failed to get it. If the answer is no, the company was pursuing alignment and succeeded.
Three structural fixes
The companies that have meaningfully reduced decision latency without sacrificing decision quality tend to have done three things.
The first is naming the decider. For every decision of consequence, one individual is designated as the decision owner. That person takes inputs from others, but owns the call. There is no appeals process short of the CEO. Once the call is made, the organization executes. This sounds trivial. In practice, most companies run on implicit deciders, which means they run on whoever is loudest, most senior, or most willing to keep the debate open.
The second is time-boxing. A decision that requires more than a defined period to make is escalated, not extended. If a cross-functional decision has not been made in six weeks, the question moves up a level. The purpose is not to rush. The purpose is to prevent the default behavior of indefinite deferral.
The third is a decision log. Decisions are recorded, along with the decider, the date, the rationale, and the expected outcome. Six months later, the log is reviewed. This produces something most companies do not have: a factual record of how their decision process performed, independent of how any given decision felt at the time.
The cost of consensus is a shadow cost. It does not appear until years later, when a competitor who did not bother with consensus has eaten the market. By that point, there is nothing to align on.