general

Bias for Action: Meaning, Examples and Interview Questions

Bias for action means choosing to decide and move with the information you have, instead of waiting for certainty that will not arrive in time. At Amazon it is the ninth of the 16 leadership principles, and it is the one that explains how a company that size still ships before its competitors do.

The meaning is narrower than "move fast". Amazon pairs speed with a test: is this decision reversible? If it is, decide now and correct later. If it is not, slow down. Candidates who miss that second half tell stories about rushing, and interviewers score them as reckless rather than decisive.

A note on a different term: "action bias" in psychology describes the urge to do something even when waiting would be better, such as a goalkeeper diving on a penalty when staying put has better odds. Amazon's principle is the opposite of that reflex. It asks for calculated speed, not motion for its own sake.

This guide covers the official wording, what the principle looks like day to day, examples at work, 15 interview questions, a worked STAR answer, and the follow-ups you should expect.

The official wording from Amazon

Speed matters in business.

That is the first sentence of the principle on Amazon's leadership principles page. The rest says that many decisions and actions are reversible and do not need extensive study, and that Amazon values calculated risk taking.

Two words carry the weight: reversible and calculated. Jeff Bezos spelled out the first in his 2015 letter to shareholders with the idea of one-way and two-way doors: a two-way door decision can be walked back, so it should be made quickly by individuals or small groups, while a one-way door deserves deliberation. In the 2016 letter he put a number on the second: "Most decisions should probably be made with somewhere around 70% of the information you wish you had. If you wait for 90%, in most cases, you're probably being slow."

What bias for action means in practice

You classify the decision before you make it. Reversible: decide today. Irreversible: gather the missing data, then decide, and say when.
You ship the smallest version that answers the question. A one-week pilot with one customer beats a one-quarter plan with a steering committee.
You set a deadline for the decision itself, not only for the work. "We pick a vendor by Thursday" prevents the meeting that spawns three more meetings.
You act on behalf of the team when the owner is unavailable, then tell them what you did and why.
You accept being wrong in public as the cost of speed, and you fix it fast when it happens.
You stop debates with an experiment. When two engineers argue about an approach, the fastest path is often to try both for a day.

Bias for action examples at work

The engineer who shipped the feature flag first. A payments team has argued for two sprints about whether a new retry policy will reduce failed transactions. One engineer implements the policy behind a flag, turns it on for 5 percent of traffic for 48 hours, and brings the numbers to the next meeting. The argument ends because there is data. The decision was reversible, so she did not wait for consensus.

The data scientist who picked a baseline. A team wants a forecasting model but cannot agree on the architecture. The data scientist ships a simple moving average into the dashboard on day two, so the business has a number to use, and improves it over the following month. The simple model turns out to be good enough for half the use cases.

The product manager who called the customer. A PM receives a vague complaint through sales that "onboarding is confusing". Instead of scheduling research, he calls three customers the same afternoon, watches one share their screen, and files two bugs and one copy change by the end of the day. The research study still happens later; the obvious fixes did not wait for it.

The operations manager who moved the shift. A warehouse operations manager sees that Tuesday afternoons are consistently understaffed. She moves four people's shifts for two weeks as a trial, tells the team it is reversible, and measures the backlog. The change sticks because it worked, and nobody waited for the quarterly staffing review.

Bias for action interview questions

Interviewers may not say the words "bias for action". They ask for a decision and then probe how you made it and what it cost. Prepare for these:

Direct questions

Tell me about a time you had to make a decision without enough information.
Describe a situation where you took a calculated risk. How did you calculate it?
Give me an example of a time you took action when it was not expected of you.
Tell me about a time you decided to move forward without consulting your manager. What was the outcome?
When have you stepped up to lead something because nobody else was doing it?
Tell me about a time speed mattered more than being thorough. How did you decide?
Describe a time you launched something before it was fully ready.

Questions that test it indirectly

Tell me about a time your team was stuck in analysis. What did you do?
Have you ever faced a crisis at work? Walk me through the first hour.
Tell me about a decision you made quickly that turned out to be wrong. What did you do next?
Describe a project where you had to learn something outside your area to keep it moving.
Tell me about a time you were frustrated by a team's lack of initiative. What did you do about it?
How do you decide when a decision needs more data and when it does not?
Tell me about a time you pushed back on a request to study a problem further.
When have you gone out of your way for a customer without being asked?

A sample STAR answer

Question: tell me about a time you had to make a decision without enough information.

Situation. I was the on-call engineer for a search service when error rates in one region jumped from 0.1 percent to 4 percent at 2 a.m. The dashboards showed a dependency timing out, but the dependency team's on-call did not respond, and the last deploy on our side had been 20 hours earlier.

Task. I had to get errors down before the European morning peak in about three hours, without being able to confirm the root cause or reach the other team.

Action. I wrote down the two options: fail over the region to a secondary cluster, which we had tested once and which would cost about 15 minutes of degraded latency, or keep waiting for the dependency team. The failover was reversible; I could flip it back within minutes. I checked that the secondary cluster had capacity, told the incident channel exactly what I was about to do and when I would reverse it if it did not help, and flipped the traffic. Errors dropped to 0.3 percent within ten minutes. I then opened a ticket with the timeline for the dependency team and kept the failover in place until they confirmed a fix at 9 a.m.

Result. The peak passed with no customer-facing incident. The dependency team found a certificate expiry on their side. In the review I proposed that regional failover become a documented first step for any dependency timeout, and the runbook was updated the same week. The next similar incident, three months later, was resolved in eight minutes by a different engineer following that runbook.

Why this answer works

It shows the calculation: the decision was reversible, the cost was bounded, and the alternative was waiting with no end in sight.
The action includes communication before and after, which is what separates a bias for action from a lone wolf.
The result is measured twice: the incident itself, and the runbook that made the next one faster.

What the interviewer probes next

"What information did you not have, specifically?" They want to see that you knew what you were missing and judged it acceptable, not that you forgot to look.
"What would have happened if you were wrong?" If you cannot answer, the risk was not calculated.
"Who did you tell, and when?" Acting alone and silently is a red flag at Amazon. Acting then informing is the expected pattern.
"Was this reversible?" Have a clear yes or no, and if no, be ready to explain why you still moved.
"What did you do differently the next time?" A bias for action includes learning from the fast decisions that went wrong.

Common mistakes

Telling a story about rushing. Skipping a security review to hit a date is not bias for action; it is a story the interviewer will score against you under Insist on the Highest Standards.
Leaving out the alternative. If the interviewer cannot see what waiting would have cost, your speed looks like impatience.
Picking a trivial decision. Choosing a library for a side project does not test the principle. Pick a decision with real stakes and real uncertainty.
Never mentioning a cost. Fast decisions have costs. Naming the one you accepted makes the story credible.
Confusing it with Deliver Results. The interviewer is scoring how you decided under uncertainty, not that the project landed.
Forgetting to close the loop. Amazon wants the decision, the result, and the correction if one was needed.

Related leadership principles

Deliver Results is where speed has to land; a fast decision that does not ship is just a meeting that ended early.
Are Right, A Lot is the counterweight: interviewers want fast decisions that are also usually correct, and candidates who seek disconfirming data before they move.
Ownership covers acting when it is not your job, which is where many bias for action stories begin.

The full list, with questions for each principle, is in our Amazon leadership principles guide. Structure every story with the STAR method and keep them in a story bank so you can pull the right one under pressure.

Want someone who has run Amazon loops to pressure-test your decisions under uncertainty? Get matched with a coach, then read Deliver Results and Ownership next.