Decide at 70% information — waiting for 90%+ almost always costs more than the marginal decision improvement returns

Why This Is a Rule

Jeff Bezos's 70% threshold (from his 2016 shareholder letter) addresses the asymmetry between information cost and information value. The first 70% of information is relatively cheap to gather and dramatically improves decision quality. The last 30% is expensive to gather (requires extensive research, rare data, or simply waiting for events to unfold) and marginally improves quality. At 70%, you're past the steep part of the learning curve and on the flat plateau.

The cost equation makes this clear. Waiting from 70% to 90% information might take weeks or months. During that wait, you're paying opportunity costs (delayed action, missed timing windows, competitor movement) and the information environment itself is changing (the 70% you already have may be degrading). Meanwhile, the decision quality improvement from 70% to 90% is marginal — you're adding precision to an answer that was already approximately right.

This complements For reversible decisions, act at 50-60% information — experiential learning from doing exceeds deliberation gains (50-60% for reversible decisions) by providing a general threshold applicable to all decisions including moderately irreversible ones. For purely reversible decisions, you can act even earlier; for irreversible ones, 70% is the minimum.

When This Fires

• When you've been researching a decision and need to determine whether to keep gathering or decide
• When analysis paralysis is setting in and you're seeking "just a bit more information"
• As a complement to Classify every decision as one-way or two-way door before deliberating — minutes for reversible, days for irreversible (door classification) for setting decision thresholds
• When someone asks "shouldn't we get more data before deciding?"

Common Failure Mode

Misidentifying 70% as 40%: "I feel like I have enough information" when you actually haven't considered major factors. The 70% threshold requires honest assessment of how much of the relevant information landscape you've covered. Feeling confident is not the same as having 70% coverage — overconfidence (After 30+ journal entries, calculate your calibration — do your 70% predictions come true 70% of the time?) can make 40% feel like 70%. Check: have you considered the major alternatives, key risks, and primary stakeholders?

The Protocol

(1) List the information you'd ideally want for this decision. (2) Assess: roughly what percentage of that ideal information do you currently have? Be honest — don't confuse confidence with coverage. (3) If ≥70% → decide. The marginal information isn't worth the delay. Set a deadline (Two-way door decisions with sub-week reversal costs get 24 hours max — emotional weight is not decision weight) and commit. (4) If <70% → identify the 2-3 most important missing information items and set a time-boxed research period to gather them. Not all missing information — just the items that would most change your decision if different from expected. (5) After the time-boxed research, reassess. If you've reached 70% → decide. If research hit diminishing returns before 70% → the remaining information may be unknowable in advance. Decide anyway and learn from outcomes.