🎯 Goal: Decide fast or slow at the right time by telling REVERSIBLE from IRREVERSIBLE decisions, using the "70% information" rule, communicating transparently, and running meetings that end with a conclusion.
Leading a startup means deciding with incomplete information — wait for full data and the chance is gone. Jeff Bezos’s trick: sort decisions into two kinds. Reversible ones ("two-way doors") should be decided fast — if wrong, fix it. Irreversible ones ("one-way doors") deserve careful deliberation. Alongside this, a good leader communicates transparently (states the reason so the team aligns instead of speculating) and runs effective meetings (clear goal, right people, a conclusion with an owner). This lesson gives you the sorting framework, the "70%" rule, and a decision template to use right away.
Lesson content
🎯
A common mistake is treating EVERY decision as if it were permanent — leading to paralysis, endless meetings with no conclusion, chances slipping away. Before each decision, ask: "If this is wrong, can I undo it, and at what cost?". Changing a button color, trying a marketing channel, posting a new type of content — reversible, so just do it and measure. But firing someone, signing a 3-year exclusive contract, betting most of your cash on one gamble — hard to reverse, so deliberate carefully and seek input. Move fast on the easy-to-fix so you have time to go slow on the hard-to-fix.
🧭
A 4-step process for deciding under uncertainty: 1) Sort the door: is this a "two-way door" (reversible) or a "one-way door" (hard to pull back)? 2) Apply the "70%" rule: for two-way doors, decide once you have ~70% of the information — waiting for 90% is usually too late. For one-way doors, gather more carefully. 3) Set an "exit trigger": for big decisions, pre-define "if metric Y isn’t met after X weeks, we change course" — turning a gamble into a controlled experiment. 4) Communicate & delegate: state clearly who decides, why, who does what, by when. A decision with no owner is an unfinished decision.
🧰
TOOL 1 — Decision-sorting framework:
Decision (example)
Reversible?
Risk if wrong
How to handle it
Change button color/copy, try 1 ad channel
Yes (two-way door)
Low
Decide fast (~70% info), measure, change if wrong
Hire/fire a key team member
Hard
High
Deliberate, consult, prepare data
Sign a long-term exclusive deal / raise money
No (one-way door)
Very high
Be very careful, ask advisors/lawyers, model the downside
TOOL 2 — Decision log template — write 6 lines for every big decision:
Field
What to fill in
Decision
What we chose to do
Door type
Two-way / one-way
Reason
Why we chose this direction
Key assumption
What we’re betting is true
Exit trigger
Which metric, if unmet, makes us change
Owner / by when
Person responsible + deadline
Logging helps you learn from decisions: months later, look back to see which assumption was wrong.
🌏
Case study — SEA digital founders deciding under uncertainty (step by step): 1) Two-way doors: test fast: new features go out to a small user group (low risk, reversible), measure the numbers, then scale. 2) One-way doors: slow down: which terms to raise on, choosing a strategic partner, entering a new market — heavy thought because they’re hard to pull back. 3) Set an exit trigger: try a new channel for 4–6 weeks on a capped budget; if it misses, stop — no throwing good money after bad. 4) Communicate transparently: explain "why we chose this direction" so the team aligns instead of speculating. (Each company’s details differ; this illustrates the principle.)
⚠️
5 traps when deciding under uncertainty: • Paralysis by perfectionism: waiting for 100% information on a reversible decision — the chance slips away. • Treating a two-way door as one-way: endless meetings over something you can fix in minutes. • Treating a one-way door as two-way: rushing to sign a big contract because "it looks like a deal", without modeling the downside. • Deciding then not communicating: the team doesn’t understand the reason, breeding speculation and lost trust. • Meetings with no conclusion: everyone talks then disperses, nobody decides or owns it — the decision "evaporates".
✅
Checklist before locking a decision: ① Is this a two-way or one-way door? ② If wrong, can I fix it and at what cost? ③ For a two-way door: do I have ~70% of the info — if so, decide now. ④ For a one-way door: have I consulted advisors and modeled the worst case? ⑤ Have I set an exit trigger (metric & deadline to change course)? ⑥ Have I stated who does what, by when, and explained the reason to the team?
Practice exercise
🔬 APPLIED EXERCISE: Write down 5 decisions you (or your project) currently face. For each: (1) mark "two-way" or "one-way"; (2) note the risk if wrong; (3) decide which to do NOW (fast) and which to pause and deliberate. Then pick one big decision and fill in the full 6-line Decision log (including an exit trigger), and write 3 sentences transparently explaining to the team why you chose that direction.
Worked template: FILLED EXAMPLE: "Try a new type of content on a new platform" → two-way door, low risk → do it now, measure for 4 weeks, change if it misses. "Drop out of school to do the startup full-time" → one-way door, high risk → pause, consult mentors and family, model the downside (what if there’s no revenue after 12 months?) before deciding.
Quick quiz
1. For a REVERSIBLE decision (low risk), the good approach is:
→ Decide fast, act, measure, and fix if wrong
A reversible decision (low risk) should be decided fast, done and measured, fixed if wrong.
2. Why distinguish reversible from irreversible decisions?
→ To spend deliberation time on the hard-to-undo ones and move fast on easy-to-fix ones
Sorting them lets you spend deliberation time on the hard-to-undo kind and move fast on easy-to-fix ones.
3. Which is an example of a HARD-to-reverse decision needing careful thought?
→ Signing a long-term exclusive deal with a partner
Signing a long-term exclusive deal is a hard-to-reverse decision that needs careful thought.
4. How does a leader’s transparent communication help?
→ It helps the whole team understand the REASON for a decision so they align instead of speculating
Transparent communication helps the team understand the reason and align instead of speculating.
5. What does the "70% information" rule for two-way-door decisions mean?
→ Decide once you have about 70% of the info, because waiting for 90% is usually too late
The "70%" rule means decide once you have about 70% of the info on two-way doors, because waiting for 90% is usually too late.
6. What is an "exit trigger" in a big decision for?
→ Pre-defining a metric & deadline: if unmet, change course — turning a gamble into a controlled experiment
Setting an "exit trigger" (metric & deadline to change course) turns a gamble into a controlled experiment.
Advanced
A deeper framework
At an advanced level, separate decision quality from outcome. A good decision is one made with sound process given the information at the time (sort the door, estimate the risk, set an exit trigger) — even if the outcome turns out bad due to luck. Great leaders judge themselves and their team by process, not just outcome; otherwise they punish sound bets that got unlucky and reward reckless ones that got lucky.
This pairs with the discipline of meetings that conclude: each meeting needs a question to answer, the right people present, and an ending of "decision + owner + deadline". For high-risk one-way doors, a useful technique is the "pre-mortem": assume it has failed 6 months from now, then ask "why?" — this surfaces risks that initial excitement hides, before you step through the door.
Judge by process, not just outcome (illustrative)
Two-way door, decided fast at 70% info, exit trigger set → bad outcome
GOOD decision (sound process, risk controlled)
One-way door, signed hastily because "it looked like a deal" → happened to profit
BAD decision (the process would be dangerous; just lucky this time)
One-way door, ran a pre-mortem, consulted advisors → good outcome
GOOD decision and good outcome
Don’t be fooled by a single outcome. Repeating a sound process wins long-term; winning by luck loses next time.
Common trap: Outcome bias: praising/blaming a decision purely on a single result, ignoring what information existed at decision time. This teaches the team to cut corners and gamble when lucky, and to dodge sound bets when unlucky — degrading decision-making over the long run.
Advanced questions
1. The team has met 3 times and still can’t settle on which headline to use for this week’s test ad. By the "reversible/irreversible" framework, where’s the problem?
→ Treating a REVERSIBLE decision as if it were permanent — decide fast, run the test, then measure
An ad headline is a "two-way door": if wrong, change it in minutes, low risk. Endlessly meeting over a reversible decision wastes time and causes paralysis. The right move: pick fast (~70% info), run an A/B test, let data answer.
2. A founder is about to sign a 3-year exclusive contract with a distributor in exchange for an attractive advance. What’s the right approach?
→ Treat it as a high-risk "one-way door": deliberate, run a pre-mortem, consult advisors/lawyers before signing
A long-term exclusive contract is hard to undo and heavily binding — exactly a "one-way door". An attractive advance doesn’t change its high-risk nature. For this kind, slow down, imagine the failure (pre-mortem), and get expert input before stepping through.
3. A decision made with sound process (sorted the door, estimated risk, set an exit trigger) turns out badly due to luck. How should you judge it?
→ It’s still a GOOD decision on process; don’t let one outcome negate a sound method
Separate decision quality from outcome. A sound bet given the information can still get unlucky. Punishing it for the outcome (outcome bias) teaches the team to dodge reasonable risk and cut corners when lucky — harmful long-term.
🎯 Real-life mission
REAL-LIFE MISSION: List 6 decisions you currently face (study, project, personal). For each, mark "two-way" or "one-way" and the risk if wrong. Pick 2 two-way ones to DO NOW this week (then measure), and 1 one-way one to pause: fill in the 6-line Decision log (with an exit trigger), run a short "pre-mortem", and get input from at least 2 people before deciding.