Jeff Bezos popularised the 70% rule: if you have 70% of the information you wish you had, you have enough. If you wait for 90%, you are already too late, because the cost of indecision compounds invisibly while you keep researching.
The subtle part of the rule is what the last 30% actually costs. Information about a decision does not arrive at a constant rate: the first 70% comes from work you control (asking, reading, estimating), and much of the remaining 30% only arrives from the decision having been made, because reality is the one source you cannot interview in advance. Waiting for 90% is often not patience; it is asking research to produce what only contact can.
Why the math favours speed
Every operating decision has two costs: the cost of being wrong, and the cost of waiting. Smart teams obsess over the first and ignore the second, which is a mistake because the cost of waiting is almost always larger than they think. A two-week delay on a launch is two weeks of opportunity cost, plus the second-order cost of every dependent decision that also slips.
The asymmetry in visibility is what makes this a leadership problem: a wrong decision produces a visible artifact with a name attached, while a slow decision produces nothing anyone can point at, just a quarter that somehow achieved less. Organisations punish what they can see, so people optimise for never being visibly wrong, which is optimising for being invisibly slow. The countermeasure is to make delay visible: when a decision stalls, write down what it is waiting for and what the wait costs per week. Most stalls do not survive being written down.
The discipline of reversibility
Decisions are either one-way doors or two-way doors. One-way doors deserve deliberation. Two-way doors do not. If you can change your mind in a week with a small cost, decide now and adjust on contact with reality. Most operating decisions are two-way doors that get treated like one-way doors.
Two refinements keep the doors honest. First, reversibility is partly a design choice: many one-way doors can be re-engineered into two-way doors by shrinking the blast radius (a rollout to 10% of users, a pilot contract instead of an annual one, a feature flag instead of a migration), and spending an hour making a decision reversible is usually cheaper than spending three weeks deliberating it. Second, watch for the disguised one-way door: choices that are technically reversible but socially sticky, like a pricing change customers anchor on. The test is not "can we undo it" but "will we, realistically, be able to."
How to actually do it
When a decision arrives, ask three questions in order. Is this reversible? If yes, decide today. What is the worst plausible outcome? If you can live with it, decide today. What new information would a delay buy? If the honest answer is "more of the same," decide today.
Then close the loop that makes 70% safe: a decided-at-70% choice needs a named owner watching the outcome and a pre-agreed signal for revisiting, so "decide fast" never degrades into "decide and forget." Teams that log their two-way-door decisions with a one-line rationale (see default to writing) discover the reassuring pattern in the data: the fast decisions get reversed rarely, and cheaply, while the slow ones were mostly slow for nothing. Confidence in the rule is built from your own log, not from the quote.