How to structure interview answers: STAR vs R-A-S

If you cannot name a number, you do not have an answer. STAR will not help you. Neither will a script you asked a chatbot to "make more senior". The interviewer is still trying not to hire the wrong person. A method is packaging. If there is no number inside, they hear a duty list with better grammar.

Same test as the CV. Vague ("I owned the dashboard", "I worked on performance", "I helped the team ship") raises risk. A result they can poke at lowers it.

The number is the answer. Everything else is order.

You need something a stranger can verify in one follow-up: time, money, errors, load, cycle time, a count of people or systems, before and after. "Improved the pipeline" is a job description. "Cut the deploy from forty minutes to four" is a claim.

If a number is already sitting in your head, skip to how you say it. If it is not, do not google frameworks. Sit down with the actual job.

If nothing comes to mind, that is a work problem

Two cases. They are not the same.

You have not mined the work yet. Most people have not. They remember tickets, standups, and tools. They do not remember what changed. Take one role, one year, and write down what was true before you touched it and what was true after:

• How long did it take. How long does it take now.
• How often it broke. How often it breaks now.
• How many people waited on it. How many still do.
• What it cost. What it costs.
• How many you handled in a week, a sprint, a quarter.

Jira, git history, a dashboard you still have access to, a Slack search, a former teammate. The number is usually in there. It does not have to be glamorous. "On-call pages dropped from a few a night to maybe one a week after we fixed the retry storm" is enough. "Review time went from two days to the same afternoon when we split the monolith PRs" is enough.

You did spend the year on theatre. Status meetings, tickets that moved because someone else wrote them, a rewrite that never shipped, "supporting the team". Then a framework is not your problem. You will not talk your way into a result you never produced. You still have to sit with the year and squeeze a proxy out of it: time you took off someone else's plate, a fire you actually put out, a count of what went through your hands. If after that honest pass there is still nothing, the news is worse than a weak interview. The work did not leave a trace, and the next hiring manager will notice the same hole.

Do not invent a percentage to fill the silence. "Improved conversion by 37%" with no source is a trap you will walk into on the second question. Round a real before/after. "Roughly 2x" from 1.5% to 3% is a judgment call you can defend. A made-up 37% is not.

If the role has no clean business metric (design, research, support, a lot of platform work), use a proxy: time saved, error rate, cycle time, volume, wait time. A proxy you can explain is better than a vanity number you cannot.

Once you have the number, say it first

Now the method matters. STAR (Situation, Task, Action, Result) puts the only useful sentence at the end. The interviewer can check out in the setup. You have already seen this on the CV: the duty line dies, the result line survives.

R-A-S is the same facts in the order people actually listen:

1. Result: the number, in the first few seconds
2. Action: what you did, one or two moves, not a tour of the stack
3. Situation: context only if they ask, or in one clause if the result is otherwise meaningless

The line that dies in the room:

So, we had this analytics dashboard, and I was responsible for it, and reporting was quite manual...

The line that survives:

I cut reporting from days to hours. I replaced a manual SQL pipeline with a self-serve dashboard. We were drowning in ad-hoc asks from the same three teams.

Same job. The second one is a claim they can test. Spoken formula: X happened, because I did Y, by doing Z. Start with X.

A fuller pass:

I cut customer churn 22% in two quarters. Sixty percent of it was in the first fourteen days, so I rebuilt activation around three behavioral milestones. We had just raised a B round. Retention was the thing the board was watching.

You do not need the board sentence unless they ask why it mattered. The first two sentences already did the work.

When STAR is the right costume

Structured behavioral rounds with a rubric (a lot of FAANG, some fintech and pharma) score Situation / Task / Action / Result in that order. There, play their game. Fill every box. Still put a number in the Result, or they have nothing to mark.

A whiteboard, a live coding pass, a system-design hour: the answer is the work on the screen. A story method does not save a wrong design.

A short checklist before the call

• You can name a before and after for every story you plan to tell.
• The number came from the work, not from a template.
• If nothing came to mind, you sat with the year before you sat with a framework.
• You can survive two "how, exactly?" questions on every figure.
• You say the result first. Setup is optional.
• You did not invent a percentage to look quantitative.

The method is there so the number gets heard. It will not invent the number.