Overview
Most bad decisions aren't caused by a lack of intelligence but by predictable traps: acting on the wrong framing, being swept up by emotion or a vivid story, and failing to think past the immediate result. A good decision process protects you from all three.
This guide combines a handful of tools into a repeatable routine you can run on anything from a hire to a strategy call.
The steps
Frame the decision correctly
Before choosing, make sure you're solving the right problem. Use abstraction laddering to check you're at the right altitude, and confirm the choice is inside your circle of competence.
Separate the modes of thinking
Emotions, facts, risks and ideas shouldn't compete at once. Run Six Thinking Hats so each gets its turn instead of the loudest winning.
Check the odds, not the story
Anchor on base rates before the compelling narrative, and watch for confirmation bias steering you toward the evidence you like.
Think two steps ahead
Ask 'and then what?' using second-order thinking, and stress-test the plan with inversion — how would this fail?
Build in a margin of safety
Assume your estimate is partly wrong and leave a buffer, per margin of safety, sized to the stakes.
Decide, and match effort to confidence
When options are truly on a par, use the hard choice model — commit to what you can stand behind rather than waiting for more data.
Key takeaways
- Framing errors cause more bad decisions than analysis errors — check the problem first.
- Anchor on base rates before the vivid story.
- Always ask 'and then what?' and 'how would this fail?' before committing.
- Leave a buffer for the times your judgement is wrong.
Frequently asked questions
What's the single biggest cause of bad decisions?
Poor framing — solving the wrong version of the problem. Get the framing right and the rest of the process has something worth optimising.
How much analysis is enough?
Match it to the stakes and your confidence. For reversible, low-stakes calls, decide fast; reserve deep analysis for the consequential, hard-to-undo ones.