Explaining a decision
Telling a past decision as a choice: the options you had, the constraint that shaped it, why you picked one and what you gave up. It shows your judgment in a way the result on its own doesn't.
What it is
A decision is a point where more than one path was open and you took one. Explaining it means giving the listener enough to judge the pick, not just to hear it: what forces were in play, what the other paths were, why you went this way, and what followed. Michael Nygard's format for architecture decision records, written for software teams, lays out the same parts: a context section that describes the forces at work and names where they pull against each other, a decision section stated in the active voice, and a consequences section that lists all of them, the negative and neutral ones as well as the positive. An interview answer about a decision is that record, spoken, in about a minute. Amazon's interview guide makes the same request from the hiring side: it asks candidates to use “I” statements so that the interviewer can see the candidate's own contribution and how they made decisions.
| Part | What to say | Why it helps the listener |
|---|---|---|
The decision | One sentence, as a choice: “I chose X rather than Y.” | It marks that a decision happened at all. “We ended up with X” sounds like weather. |
The constraint | The time, budget, headcount, data, platform or rule that narrowed the field. | It explains why the obvious option wasn't open, so the listener can judge the choice against the real field. One published set of leadership principles puts it as constraints breeding resourcefulness. |
The options | At least one other path that was credible, and in one clause what it offered. | A choice with no real alternative is an event. The alternative shows you saw the field. |
The reason | Why this option won, tied to the constraint or the goal. | This is the judgment the listener is assessing. “Best practice” names no reason. |
The cost | What it gave up or made harder, and what you did to bound that. | Naming the cost shows you saw the full consequences, the negative ones included, as a decision record should. A decision described with no downside leaves the listener wondering what you missed. |
Who decided | Your call, a recommendation someone else approved, or a call you disagreed with and then backed. | It keeps the ownership honest and tells the listener how much of the judgment is yours. |
Say who decided. Some decisions you'll want to talk about were recommendations: you did the thinking and someone with the authority signed off. That still counts, as long as you say so. The worked example in Amazon's own article on its Leadership Principles has the candidate gathering feedback from twenty customers and recommending a two-week delay to a launch; the thinking it shows is the candidate's whether or not the final approval was. The reverse case counts too. In the 2016 shareholder letter, Jeff Bezos describes telling a team he disagreed with a project and committing to it anyway, and calls that something a boss should do as well as a report. If you argued for Y, lost, and then made X work, say that in those words: it's a decision about how to act, and the reasoning behind it is yours.
Match the depth of deliberation to the stakes. Amazon's Bias for Action principle states that many decisions are reversible and don't need extensive study. The same shareholder letter argues that most decisions should be made with around 70% of the information you'd like, that waiting for 90% is usually too slow, and that being good at correcting a wrong call makes being wrong cheaper than being slow. Our advice: say which kind of decision yours was. For a reversible one, a strong answer says why it was safe to move quickly and what signal you watched so you'd notice a wrong turn. For one that was hard to undo, a strong answer says what you checked first and whose view you sought, which is what the Are Right, A Lot principle describes as seeking diverse perspectives and working to disconfirm your own belief.
When to use it
Use this shape whenever a question asks for a decision outright: “tell me about a difficult decision”, “a time you had to choose with incomplete information”, “a trade-off you made”. Use it just as much inside a project walkthrough or a STAR story: look for the step in your Action beat where you had options, and when there is one, that step is where to slow down and give the parts above. Google's re:Work guide to structured interviewing says its questions come with predetermined follow-ups designed to draw out detail about the candidate's approach, and that interviewers may be judging the thought process rather than a correct answer. Laying out the constraint, the options and the reason before the follow-up asks for them uses that time on your terms.
When not to
The full shape is not right for everything that looks like a decision.
- A hypothetical. “What would you do if…” asks how you would decide, not what you decided; the re:Work guide treats hypothetical and behavioral questions as different kinds. Walk through what you'd find out first and how you'd weigh it, and offer a past decision briefly as evidence if one fits. See Answering a hypothetical.
- A step with one workable path. If there was no real alternative, call it what it was and move on. Inventing a second option to make it look like a choice tends to fall apart at the first follow-up, and it spends time you could give to a real decision from the same project.
- A small, reversible call. The Bias for Action principle quoted above holds that many reversible decisions don't need extensive study. Narrating a long deliberation over one of those suggests you deliberate over everything. A sentence on why it was safe to try and how you'd back it out is enough.
- A narrow follow-up. If the interviewer asks just what the cost was, give the cost. Restarting from the constraint spends their time on parts they've already heard.
- A call you had no voice in. If a decision was made above you and handed down, say so plainly and talk about how you carried it out. Claiming it as yours is a misrepresentation, and one a follow-up about the alternatives you weighed can expose.
Compared
| Part | Weak | Strong | When the short form is fine |
|---|---|---|---|
Decision | “We ended up sending the emails in the background.” | “I chose to send confirmation emails from a background job instead of inside the checkout request.” | When that decision is scenery for a story about a different one. |
Constraint | “There were a lot of constraints.” | “We had two engineers, six weeks to launch, and no budget to run a new service.” | When the interviewer heard the setting in an earlier answer; a half sentence to recall it. |
Options | “There wasn't really another way.” | “The alternative was to retry inline with a short backoff. It was simpler, but it held the request open while the mail provider was slow.” | When the alternative was plainly worse and you've said why in a clause. |
Reason | “It was best practice.” | “Checkout speed was the goal of the project, so I took the option that kept the request short even though it added a moving part.” | Rarely. The reason is the part the listener is assessing. |
Cost | “It worked great.” | “Emails now arrive up to a minute late, and we have one more job to operate. I capped the delay and alerted on the backlog.” | When the call was small and reversible: one sentence on how you'd undo it. |
Who decided | “We decided.” | “It was my proposal. My lead approved it on the condition that we added the backlog alert.” | When it was plainly your own call and you've been saying “I” throughout. |
In a picture
How to say it in an interview
The common mistake
Quiz
Related
Sources
- Documenting Architecture Decisions (Michael Nygard, Cognitect)
- Your complete guide to the Amazon interview process
- Amazon's Leadership Principles: Interview strategies from a manager
- Amazon Leadership Principles: Values and company culture (About Amazon)
- Jeff Bezos' 2016 Letter to Amazon Shareholders
- A guide to structured interviewing for better hiring practices (Google re:Work)