Bias for Action is the discipline of defaulting to motion. Most decisions are reversible, so the cost of being wrong is usually a quick correction, not a catastrophe. The model says: act on roughly 70% of the information you wish you had, rather than stalling until you reach 90%. That last 20% almost never arrives cleanly, and the weeks spent chasing it are weeks the market never gets to react to.
It works because reality is a faster teacher than analysis. A landing page launched today with placeholder copy generates real signal: clicks, confusion, silence. A week of refining messaging nobody has seen generates only opinions. Action converts uncertainty into data you can actually act on.
In practice, the trigger is catching yourself researching instead of testing. Ask one question: is this decision reversible? If yes, move now and correct later. Ship the rough version, send the imperfect email, run the small test. You buy the right to be wrong cheaply, and you compound learning while slower competitors are still gathering inputs.
Common mistake
Applying it to irreversible, high-stakes decisions where speed is reckless. Bias for Action is a default for reversible calls; the reversibility test is the whole point, and skipping it turns a useful heuristic into an excuse to act rashly.
How to use it
Three ways to put it to work.
Everyday life
Stuck choosing the right preschool for your kid
You have toured three preschools and keep requesting more data - waitlist odds, lunch menus, a fourth tour. Ask: is this reversible? Yes, you can switch schools mid-year if it flops, so you do not need 90 percent certainty. Two of the three feel right at 70 percent, so put down the deposit on the closer one today and revisit in November if it is not working.
Tech
Debating which database before you have any users
Your team has spent two weeks comparing Postgres versus a fancier distributed store for an app with zero traffic. Ask: is this reversible? Yes, a migration is a weekend of work at this scale, not a one-way door. Stop the bake-off, default to boring Postgres at 70 percent confidence, and ship the feature this sprint - you can re-platform later if real load ever proves you wrong.
Small business
Cafe owner agonizing over a new oat-milk latte price
You have been modeling margins for a week, unsure whether to charge 5.50 or 5.75 for the new latte. Ask: is this reversible? Yes, a chalkboard price changes in ten seconds. Pick 5.50 now on 70 percent of the data, print it on the board today, and watch a week of sales - if it moves volume but kills margin, bump it to 5.75 next Monday.
Reckon · Thinking tool
Bias for Action
Amazon Leadership Principles, rooted in military doctrine
What it is
Most decisions are reversible. Default to acting on 70% information rather than waiting for 90%.
When to use it
When you catch yourself researching instead of testing. Ask: is this decision reversible? If yes, move now and correct later.
Work through it.
Print this (use the Save as PDF button up top) and fill it in on paper. Identify one decision you have been researching past the point of usefulness and move on 70% information right now.
01
THE STALL: Name the specific decision or action you keep putting off because you want more information or certainty.
02
THE REVERSIBILITY TEST: Is this decision permanent or reversible? Write out exactly what you would do if the action turned out to be wrong.
Most stalls are reversible. Naming the undo path collapses the perceived risk.
03
THE 70% AUDIT: List what you already know. If you have at least 70% of the information needed to make a reasonable call, you have enough.
04
THE COST OF WAITING: What is the concrete cost of not deciding for another week? Write it as a number or a specific missed outcome.
05
THE ACTION: Write the specific next step you will take in the next 24 hours. Not 'research more.' The action itself.
Reckon · reckon.report · flip for the working template
Reckon · Working template
Bias for Action
Print double-sided. Read the how-to on the front; do the work here.