{"id":877,"date":"2026-10-22T18:12:00","date_gmt":"2026-10-22T16:12:00","guid":{"rendered":"https:\/\/grindloop.ai\/blog\/?p=877"},"modified":"2026-10-21T10:06:24","modified_gmt":"2026-10-21T08:06:24","slug":"improved-a-process-interview-question","status":"publish","type":"post","link":"https:\/\/grindloop.ai\/blog\/improved-a-process-interview-question\/","title":{"rendered":"Tell Me About a Time You Improved a Process: Prove the Team Still Uses It"},"content":{"rendered":"<p>An interviewer asks &#8220;tell me about a time you improved a process.&#8221; What they want to know is whether the change stuck. Most candidates describe a script or a new checklist and stop there. The interviewer then asks who still uses it. So build the answer around adoption from the start. First, give the cost of the old process as a number you measured. Second, describe the change in one or two sentences. Third, spend the most time on how you got other people to use it. Say what you tried first and why it failed. Then say how you made the new way the default. Fourth, give evidence that it is still in use. That could be a metric months later or another team that copied it. End with what the change cost. This post follows one Android engineer and a flaky test suite.<\/p>\n<h2>When you improved a process, the interviewer scores whether people kept using it<\/h2>\n<p>Amazon&#8217;s <a href=\"https:\/\/www.amazon.jobs\/content\/en\/our-workplace\/leadership-principles\" target=\"_blank\" rel=\"noopener\">Leadership Principles<\/a> include one that fits this question. The page carries no date. Under Invent and Simplify, it says leaders &#8220;always find ways to simplify.&#8221; It also warns that new ideas may be &#8220;misunderstood for long periods of time.&#8221; That second line is the hard part of any process change. You built the better way. Your teammates kept doing the old thing anyway.<\/p>\n<p>Will Larson made a related point about engineering migrations in an <a href=\"https:\/\/lethain.com\/migrations\/\" target=\"_blank\" rel=\"noopener\">April 2018 essay<\/a>. Larson is an engineering leader and author. He wrote that &#8220;starting but not finishing migrations often incurs significant technical debt.&#8221; A process change works the same way. Picture half the team on the new way and half on the old. Now there are two ways to do one job. So a tool nobody adopted is a half-finished change. The interviewer knows this. That is why the follow-up asks about adoption.<\/p>\n<h2>Marco&#8217;s first version reached two of nine engineers<\/h2>\n<p>Marco is an Android engineer at a travel booking company. His team&#8217;s UI tests failed at random on CI. Engineers hit &#8220;re-run&#8221; until the build went green. Nobody trusted a red build anymore.<\/p>\n<p>Marco wrote a script that ran a failing test 20 more times. If it passed some of those runs, the script marked it flaky. Then it opened a ticket for the test&#8217;s owner. He posted the script and a wiki page in the team channel. After a month, two of nine engineers were using it. The re-run rate barely moved.<\/p>\n<p>Many candidates stop the story at the script. It sounds like initiative. But the interviewer hears a side project with two users. The rest of the story is how Marco got the other seven.<\/p>\n<h2>Measure the old process before you change it<\/h2>\n<p>A baseline turns &#8220;it was slow&#8221; into a cost the interviewer can write down. Pick one number that the old process made worse. Then measure it before you change anything.<\/p>\n<p>Marco pulled three months of CI history. About one build in four needed at least one re-run. Each re-run added around 25 minutes. That came to roughly 30 engineer-hours a month spent waiting. He had not measured any of this before writing the script. He did it after the first version failed, because he needed a reason his lead would accept.<\/p>\n<p>Say that order honestly if it is true for you. Measuring late is common. Claiming you measured first, when you didn&#8217;t, falls apart on the first follow-up.<\/p>\n<h2>Make the new way the default so nobody has to remember it<\/h2>\n<p>Larson&#8217;s essay describes a step he calls stopping the bleeding. All new work has to use the new approach, enforced by tooling. Opt-in tools depend on people remembering them. A default does not.<\/p>\n<p>Marco moved the script into CI itself. A failing test now got its 20 retries automatically. If it was flaky, CI quarantined it and filed the ticket. The build stayed green. Nobody had to remember anything. He also added one rule for new tests. Each one had to pass 20 runs in a row before it could merge. His lead agreed to try it for two sprints. He asked for that trial, not a permanent change, because a trial was easier to say yes to.<\/p>\n<p>Name who had to agree and what you offered them. That shows you treated the change as other people&#8217;s decision too. Our post on <a href=\"https:\/\/grindloop.ai\/blog\/ownership-interview-question-whose-job\/\">the ownership interview question<\/a> covers the same habit of leaving owners in charge.<\/p>\n<h2>Marco&#8217;s flaky-test answer, about two minutes long<\/h2>\n<p>We invented Marco, his team and his numbers to show the shape of the answer. Your own story should keep the shape and use what really happened.<\/p>\n<blockquote>\n<p>Our Android UI tests failed at random on CI. About one build in four needed a re-run. Each one cost about 25 minutes. Across the team, that was roughly 30 hours a month of waiting.<\/p>\n<p>I built a script that retried a failing test 20 times. If the test passed some runs, it was marked flaky and the owner got a ticket. I shared it on the wiki. A month later, only two of nine engineers used it. People forgot it existed when a build went red.<\/p>\n<p>So I asked my lead for a two-sprint trial of a different setup. I moved the retries into CI, so every failing test got them automatically. Flaky tests were quarantined and ticketed without anyone acting. I also added a rule that new tests had to pass 20 runs before merging.<\/p>\n<p>After the trial, re-runs dropped to about one build in twenty. The team kept the setup. Six months later, the quarantine list is still in use. The web team copied the setup too. The cost was CI time. Each build got about four minutes slower because of the new-test rule.<\/p>\n<\/blockquote>\n<p>The answer gives the baseline first, so the result has something to compare against. It admits the first version failed. That failure is what makes the second version believable. And it ends on a number from months later, plus one cost.<\/p>\n<h2>&#8220;Who still uses it?&#8221; needs a fact from after you stopped pushing<\/h2>\n<p>Expect this follow-up, or a version of it. &#8220;Is it still running?&#8221; &#8220;What happened after you left the team?&#8221; The interviewer is checking whether the change depended on you.<\/p>\n<p>Answer with something you didn&#8217;t do yourself. Marco has two facts. The quarantine list still gets new entries six months later. And the web team set up the same check on their own. Neither needs him to keep promoting it. A weaker reply is &#8220;people really liked it.&#8221; That tells the interviewer the change had fans. The follow-up asked whether it ran without its author.<\/p>\n<p>Check your facts before the interview if you can. Ask a former teammate whether the tool still runs.<\/p>\n<h2>If your change was rolled back, tell that instead<\/h2>\n<p>Some honest stories lack the key element. The tool was dropped after you left, or the team never adopted it. Don&#8217;t stretch &#8220;a few people liked it&#8221; into adoption. The follow-up will expose it.<\/p>\n<p>Tell the story with the rollback in it. Say what you built, how far it got and why it stopped. Then name the step you skipped. Usually it is one of two steps. You never made the change the default. Or nobody owned it after you. Then say what you did differently on your next change. A rolled-back tool plus a lesson you later applied is a complete answer. But it now reads partly as a failure story. For that kind of story, put the decision first. Our post on <a href=\"https:\/\/grindloop.ai\/blog\/tell-me-about-a-time-you-failed-production-incident\/\">the &#8220;tell me about a time you failed&#8221; question<\/a> shows how. If you have a smaller improvement that did stick, it may be the stronger pick for this question.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Tell me about a time you improved a process? Show the baseline, how you made it the default and proof the team still uses it. A full Android sample answer.<\/p>\n","protected":false},"author":4,"featured_media":879,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"rank_math_title":"Tell Me About a Time You Improved a Process: Prove the Team Still Uses It","rank_math_description":"Tell me about a time you improved a process? Show the baseline, how you made it the default and proof the team still uses it. A full Android sample answer.","rank_math_focus_keyword":"improved a process","footnotes":""},"categories":[10],"tags":[35,16,19,18],"class_list":["post-877","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-behavioral","tag-behavioral-interview","tag-interview-prep","tag-soft-skills","tag-star-method"],"_links":{"self":[{"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/posts\/877","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/comments?post=877"}],"version-history":[{"count":1,"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/posts\/877\/revisions"}],"predecessor-version":[{"id":878,"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/posts\/877\/revisions\/878"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/media\/879"}],"wp:attachment":[{"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/media?parent=877"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/categories?post=877"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/grindloop.ai\/blog\/wp-json\/wp\/v2\/tags?post=877"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}