Skip to content

Putting numbers on impact

Stating the result of your work as a measured change: the metric, the number before, the number after, and how much of the change was yours. It lets an interviewer weigh the outcome rather than take your word for it.

On this page

What it is

Putting numbers on impact means describing the outcome of your work as a measured change rather than an activity. "I rebuilt the upload service" is an activity. "Failed-upload tickets fell from about 120 a week to about 30" is an outcome someone else could check. Amazon's interview guide calls this the "so what?" test: a strong account doesn't stop at what you did, it shows what changed because you did it, and the guide asks for metrics and data wherever they exist. Yale's career office makes the same point about resume lines, and adds a piece that is easy to skip: give a baseline, so the reader can see how big the change was against where things started.

PartWhat it answersExample (illustrative)

Success defined

What did "done well" mean before the work started?

"The goal set at kickoff was to halve upload failures by the end of the quarter."

Metric

Which number or observable outcome measures that?

"We tracked support tickets tagged upload-failed, per week."

Baseline

What was the number when you began?

"About 120 a week, averaged over the prior two months."

After

What was the number when you finished, and when did you measure it?

"About 30 a week, measured over the six weeks after launch."

Attribution

What else moved the number, and what share was yours?

"A mobile release in the same month fixed one client bug; I'd credit the service change with most of the drop, not all of it."

Without it

What would likely have happened if the work hadn't been done?

"Tickets had been flat for a year, so I don't think they'd have fallen on their own."

Why the baseline and the date matter. A number on its own ("we hit 30 tickets a week") can't be judged: 30 is a success if you started at 120 and a failure if you started at 35. The baseline turns a figure into a change. Saying when you measured the after-number matters too, because a result read the day after launch can be a dip that recovers. The field of impact evaluation makes the same distinction: the stronger study designs collect their baseline before the work starts, and a simple before-and-after reading on one group is counted as the weakest design, because other things change in the same period.

Why attribution matters. Impact, in the evaluation sense, is the difference between what happened and what would have happened without the work, and that second thing is estimated rather than observed. If your project had a control group or an A/B test, that comparison is your strongest evidence, so lead with it. When there was no comparison group, you can do the honest version: name the other things that moved the metric at the same time (a release, a season, a hiring freeze), say which part of the change you believe came from your work, and say why you believe it. One of Amazon's leadership principles, Dive Deep, asks leaders to be skeptical when metrics and anecdotes disagree. Our advice: apply that to your own number before the interviewer does. If the metric improved but the people affected didn't notice, say so and explain the gap.

Outputs are not outcomes. Evaluation frameworks separate inputs (what went in), outputs (what you produced), outcomes (what changed as a result) and longer-term impact. "Shipped 14 endpoints" and "ran 30 user interviews" are outputs. They show effort, not effect. Lead with the outcome and keep the output as supporting detail about how you got there.

When to use it

Use it wherever you close a story about your own work: the Result beat of a STAR answer, the last line of a project pitch, and any answer that describes a past project. Amazon's published guidance asks that each example end with measurable outcomes, names "not enough data" among the common candidate mistakes, and asks for specific, quantifiable detail in place of general claims while keeping each answer to about four minutes. Those are one employer's stated preferences, not a universal rubric, but the shape they describe is what a concrete result looks like anywhere.

Prepare the numbers before the interview, not during it. For each story you plan to tell, write down the metric, the before, the after, the date of the after-reading, and the one or two other things that moved it. If you've left the company, you may have to work from memory: say "roughly" and give the range you're confident in. A rough number stated as rough is more credible than a precise one you can't defend under a follow-up.

When not to

Some results don't have a number, and inventing one is worse than saying so. Even the most data-minded published interview advice allows that a result can be qualitative, and asks for specific feedback you can describe in that case. Our version of that: give an observable outcome, something a colleague could have seen. The on-call rotation stopped getting paged at night. The weekly status meeting was canceled because the dashboard replaced it. A customer who had threatened to leave renewed. These are checkable even though they aren't counts.

Other cases where a number is the wrong lead:

  • The question is about your judgment, not your result. "Tell me about a decision you'd make differently" or "how did you handle the disagreement" wants the reasoning. A result line at the end is fine; a result line in place of the reasoning isn't.
  • The work isn't finished. Say what the target is and what the early reading shows, and label it as early. Don't report a projected number as an achieved one.
  • The number isn't yours to claim. If a company-wide metric moved and your project was one of several contributors, lead with the metric you directly moved (your service's error rate) and mention the larger one as context.
  • A failure story. When the outcome was bad, the number that matters is the honest one: what it cost, how long it took to notice. See Failure and conflict without blame.

Compared

Result lineWhat it gives the listenerWhen it's the right choice

"It went well and the team was happy."

Nothing checkable.

Hard to justify. Even a qualitative result can name what changed and who noticed.

"We cut upload failures by 75%."

A change, but no base, no date, and "we" hides your part.

When the base is large and well known to the listener, and you follow with the before and after if asked.

"Tickets went from about 120 a week to about 30, measured over six weeks after launch."

A metric, a baseline, an after, and a window. The listener can judge the size.

Our default recommendation when the numbers exist and you can defend them.

"...about 30 a week. A mobile fix the same month took some of that; I'd put most of the drop on the service change, and tickets had been flat for a year before."

The line above plus attribution and the likely path without the work.

Project deep dives and senior rounds, where attribution and the path without the work are natural follow-ups. In a 30-minute phone screen, keep it to the line above and have this ready.

"The night pages stopped. Nobody measured it, but the on-call channel went quiet."

An observable outcome, honestly labeled as unmeasured.

When no metric existed. Better than a number you'd have to make up.

In a picture

Success definedMetric chosen

Step 1 of 4

Define success before the work

Start where the project started: what was the goal, in measurable terms, when it was scoped? If nobody wrote one down, say what you took the goal to be.

How to say it in an interview

The common mistake

Quiz

A candidate says: "After my change, page load was 1.2 seconds." What's the most important thing missing?
Your project launched in the same month as a marketing campaign, and sign-ups doubled. How should you report it?
Nobody measured the outcome of your work, but the nightly on-call pages stopped. What's the best result line?
Which of these is an outcome rather than an output?

Related

Sources

  1. Your complete guide to the Amazon interview process
  2. 6 common mistakes candidates make during their Amazon job interview, and tips to avoid them
  3. Amazon Leadership Principles
  4. Writing Impactful Resume Bullets (Yale Office of Career Strategy)
  5. Impact evaluation (Wikipedia)
Sign in to report a mistake