High-Velocity Decision-Making With Imperfect Information

Definition

Most decisions are made in conditions of uncertainty. Waiting for 90% of information before deciding is often slower than deciding at 70% and having the capability to rapidly recognize and correct if the decision is wrong. This principle decouples decision speed from information completeness, but with a critical condition: the organization must be genuinely good at detecting bad decisions and course-correcting quickly. High-velocity decision-making only works if you also have high-velocity error detection and correction.

In the Book

Bezos articulates this in 2015 as the second thought on high-velocity decision-making: "Most decisions should probably be made with somewhere around 70% of the information you wish you had. If you wait for 90%, in most cases, you're probably being slow. Plus, either way, you need to be good at quickly recognizing and correcting bad decisions. If you're good at course correcting, being wrong may be less costly than you think, whereas being slow is going to be expensive for sure."

The logic is economic: slow decisions have a fixed cost (the decision cycle time), and the marginal information you gather between 70% and 90% often doesn't change the answer. But waiting costs you the decision cycle, and in fast-moving environments, the environment changes while you gather information. Speed has value, and if you can course-correct, the cost of being wrong at 70% is often less than the cost of waiting.

The implicit requirement is clear: "You need to be good at quickly recognizing and correcting bad decisions." This means you need visibility into outcomes, willingness to change course, and processes that allow you to update without shame.

Why It Matters

This concept formalizes the speed-accuracy tradeoff in decision-making and permits organizations to choose to move faster if they have the capability to detect errors. It applies across domains: any organization that can see the results of decisions and change course quickly can afford to make decisions with less certainty. It inverts the common assumption that more information always leads to better decisions; instead, it asks: at what information level does the cost of gathering more information exceed the benefit? And in fast-moving domains, that threshold is often lower than most teams assume.