The biggest impact interview question still has an answer when nobody measured your work. Pick the change that reached the most people beyond you. Then rebuild its before and after from records that already exist. Crash and ANR rates, support tickets and release notes all count. So does what another team did with your work afterwards. Give each number with its source and call an estimate an estimate. The interviewer will ask how you know. A figure with a source survives that question. A round figure you guessed in the room usually does not. Also say what else changed in the same release. Then the credit you claim is credit you can defend. “It made the app much faster” can’t be checked. “Our ANR rate fell from about 0.9% to 0.3% in Play Console” can. The running example below is an Android engineer whose startup fix nobody ever put on a dashboard. The last section covers the number already printed on your résumé.
The biggest impact interview question is scored on how far your work reached
The size of your number matters less than how far the work reached. Lior Neu-ner is a former Meta engineer and engineering manager. He wrote a guest post on how Meta evaluates behavioral interviews for interviewing.io. That company sells mock interviews. The post carries no publication date. Its example answers differ by level mainly in scope. A junior answer shows impact on the team. A staff answer shows “a large impact on their org.”
Jackie Bavaro describes the same idea for product managers. Her July 2025 newsletter post covers the project-you’re-proud-of question. She sums up what interviewers want in three words: “Scope. Autonomy. Impact.” Her weak example sounds passive. It describes good work that mostly followed someone else’s plan.
A vague reach costs you more than a missing metric. Say your fix changed how two teams ship. That can beat a 40% speedup on a screen few people use. You still need evidence for the reach. That evidence is usually easier to find than a business metric.
Rebuild the before and after from records that still exist
Tomas is an Android engineer with five years of experience. He works on a podcast app and is interviewing for a senior role. We invented him for this post.
His best story has no dashboard behind it. The app froze for several seconds on first launch on cheap phones. Tomas traced it to the episode database migration and the image cache. Both ran on the main thread at startup. He moved both to a background worker and showed cached episodes first. He also added a build check. It fails when launch code touches the database on the main thread. Nobody set a goal for the fix. Product analytics never tracked startup time. So when he drafted his answer, he wrote “launch felt much faster.” That sentence gives the interviewer nothing to write down.
Then he looked for records that existed anyway. Three turned up.
The first was Google Play’s Android vitals. The Android vitals page carries no visible date. It sets a bad behavior threshold of 0.47% for the user-perceived ANR rate across devices. A single phone model has a threshold of 8%. Above a threshold, it says “Play may reduce the visibility of your title.” Play judges this on a 28-day average. In Play Console, Tomas found the app’s ANR rate at about 0.9% before his release. One budget phone model sat above 8%. Two months after the release, the overall rate was about 0.3%.
The second was the support queue. Tickets tagged “frozen” or “stuck on launch” fell from about 40 a month to about 8.
The third showed reach. The next quarter, the video team copied his build check into their own app. That one fact moves his story from his own screen to the wider org.
Carrus is a coaching marketplace that sells interview sessions. Its guide to the STAR result has no date or byline. It suggests comparing your result with the year before or with the plan. If you lack data, it suggests asking former colleagues for rough estimates. Or it says to pick a different example. If no record of the change exists anywhere, the story may be the wrong one for this question.
Say where each number came from and what else shipped with it
The same Carrus guide expects the follow-up. It says a savvy interviewer asks, “Is that a good result? How does it compare?” Tomas has two answers ready. The 0.47% threshold gives his ANR rate a scale. The support tickets give a second source that doesn’t depend on Play.
The harder follow-up is “How do you know it was your change?” Tomas’s release also included a newer image library. He says so before he is asked. Then he says why he still credits the startup fix. The ANR traces in Play Console had pointed at the migration call on the main thread. After his release, that trace dropped out of the top reports.
Label your estimates in the sentence itself. “About 40 tickets a month” tells the interviewer you rounded. “Roughly” or “about” costs nothing. A precise figure you can’t source costs a lot if they ask twice.
Put the pieces in this order when you speak. Start with the problem and who it hurt. Then say what you changed, in one or two sentences. Give the before and after, with a source for each number. Name what else shipped. End on the reach, because that is what the level is judged on.
What Tomas says when asked for his biggest impact
Tomas, his podcast app, the tickets and every figure below are made up for this post. The vitals thresholds are real. Keep his order and use your own records.
The work with the widest reach was a startup fix nobody had asked for. Our app froze for several seconds on first launch on cheaper phones. I traced it to our database migration and image cache running on the main thread.
I moved both to a background worker and showed cached episodes first. Nobody tracked startup time, so I measured the result from what we had. In Play Console, our ANR rate was about 0.9% before the release. Google’s bad behavior line is 0.47%. Two months later we were around 0.3%. Support tickets about the app freezing went from about 40 a month to about 8.
One caveat. That release also upgraded our image library. I still credit the startup change, because the top ANR trace was the migration call. It disappeared after the release.
The part I’m proudest of is what happened next. With the fix, I added a build check. It fails when launch code touches the database on the main thread. Our video team copied it into their app the next quarter.
The caveat paragraph answers the attribution question before the interviewer asks it. For more on picking the story itself, see our post on the most challenging project question. If your interviewer is drilling into a metric that looked fine, our dive deep post covers that case.
Your résumé already states a number, so make it match your answer
Your interviewer may have your résumé open during the round. Tomas’s old résumé said “improved app startup performance by 60%.” He couldn’t remember where the 60% came from. If the interviewer asks about that line, the answer above won’t help. The résumé claims 60% and his spoken answer gives an ANR rate. The interviewer now has two numbers and one of them has no source.
So Tomas rewrote the line to match his records. It now reads “Cut ANR rate from about 0.9% to 0.3% by moving startup work off-thread.” Now the résumé and the answer support each other. Check each number on your own résumé the same way. If you can’t name its source in one sentence, rewrite the line or take the number out.