Work Was Deprioritized Midway Through a Project? Answer With What You Left Behind

Asked about a time your work was deprioritized midway through a project? The interviewer is watching two things. One is how you act when someone else’s decision costs you weeks of work. The other is what you do with the half-built code. Answer in four moves. Say what you were building and how far you got. Then say how you heard, what you asked and the one case you made. Next, say how you left the unfinished work so someone could restart it. Finish with where your effort went and what became of the parked work. The third move carries the proof. “I moved on” can’t be checked. A restart note, a flag left off and a later pickup can. Below, an Android engineer builds this answer from a cart rewrite that lost to a payments deadline. The last section covers the case where committing to the new plan is the wrong move.

When your work was deprioritized, the interviewer is scoring your reaction

The project itself matters less here than in most behavioral questions. You didn’t choose to stop it. So the interviewer can’t judge the call. They judge what you did after it.

Questions about change come up in Google’s behavioral round. IGotAnOffer is a coaching site that sells mock interviews with former Google interviewers. Its Googleyness guide was last updated in June 2026. It lists “tell me about a time you had to change” among typical questions. The same guide cites a Google Careers video in which recruiters name three traits. They are comfort with ambiguity, a collaborative nature and a bias to action. It also quotes Laszlo Bock, who ran Google’s HR for over ten years. In his 2015 book Work Rules!, Bock tied comfort with ambiguity to one admission. He wrote that “we don’t know how our business will evolve.”

A deprioritized project is that admission landing on your desk. The business changed its mind halfway. The interviewer wants to know if you sulked, stalled or fought for months. They also want to know if you quietly kept going on the old plan.

Amazon writes the expected behavior down. Its Leadership Principles page carries no date. One principle is called Have Backbone; Disagree and Commit. It says leaders should respectfully challenge decisions they disagree with. Then it says “once a decision is determined, they commit wholly.” Your answer should show both halves, in that order.

Ask why once, make one case with a number, then commit

The running example is Jonah, a made-up Android engineer with six years of experience. He works on a grocery delivery app. He is interviewing for a senior role at a travel company.

His project was an offline-first cart. Shoppers often edit carts inside stores with weak signal. The old cart lived only on the server. About 3% of cart edits failed on bad connections. Jonah planned eight weeks to move the cart into a local Room database with background sync. By week five, the schema and the read path were done. The write path and the conflict rules were not.

Then his director moved the whole team onto checkout. The company had signed a new payment provider. The contract set a date for going live.

Jonah did three things that week. First, he asked what had changed and whether the date could move. It couldn’t. Second, he made one case, in writing. The finished half could ship in four days behind a remote flag. That would cut failed edits on the read side right away. Third, when the director said no, he stopped arguing. He took the hardest piece of checkout. That was handling payment errors when the provider’s confirmation screen timed out.

Make exactly one case. Making none sounds like you didn’t care about your own work. Arguing for three weeks sounds like you never committed. Put a number in the case, because a number gives the director something to weigh. Jonah’s was the 3% of edits that failed.

Expect a follow-up here. “Did you agree with the call?” Jonah’s honest answer is partly no. He still thinks the offline cart mattered. GitLab’s handbook page on making decisions gives him words for that. The page shows no date. It says the person who owns a decision “is not responsible for how people feel.” The page then asks team members to “disagree, commit, and disagree.” So Jonah can say he still disagrees and that he shipped checkout on time anyway. That sounds more credible than claiming he agreed all along.

Leave the half-built work so a stranger could restart it

“We paused it and I moved to checkout” leaves the interviewer with a question. What happened to five weeks of code? If the answer is a stale branch nobody can read, those weeks are gone.

Jonah spent his last two days on the cart doing three things. He merged the database schema into the main branch, with the remote flag off. That kept the code from drifting while checkout work landed on top. He deleted the half-done write branch, after recording what was in it. Then he wrote a one-page restart note and linked it from the backlog ticket.

The note covered four points. It listed what was done and tested. It listed what was left, mainly the rules for resolving conflicts between local and server carts. It named one risk. The schema migration had to run before anyone turned the flag on. Last, it estimated the restart cost at about three weeks.

Four months later, a product manager reopened the offline cart. Another engineer, Ana, picked it up from the note. She shipped it in just under three weeks. She needed one 30-minute call with Jonah.

That pickup is the strongest line in the story. It shows the cart work survived Jonah’s move to checkout. It also shows the note worked without its author in the room.

Jonah’s deprioritized-work answer in four paragraphs

Nothing below happened. Jonah, his grocery app and the numbers were all made up to show the four moves in order. Fill the same slots from your own history.

I was five weeks into an eight-week project to make our grocery cart work offline. About 3% of cart edits failed on weak in-store signal. The local database and read path were done. The write path and conflict rules weren’t.

Then the company signed a new payment provider with a fixed go-live date. My director moved the whole team to checkout. I asked whether the date could move. It couldn’t. I made one case in writing. The finished half could ship in four days behind a flag. He turned it down. So I took the hardest checkout piece. That was the payment timeout handling. We shipped on the contract date.

Before switching, I spent two days parking the cart work. I merged the schema with the flag off. I deleted the half-done branch after recording what was in it. Then I put together a one-page restart note. It covered what was done and what was left. It also named one risk and gave a three-week estimate.

Four months later the cart came back. Another engineer picked it up from that note and shipped it in under three weeks. I still think the cart mattered. But the note meant pausing it cost us very little.

Sometimes you are the one who drops the work. Our post on the competing priorities question covers that version.

Commit stops being the right move when half-shipped work can hurt users

Everything above assumes the unfinished work is safe to leave. Sometimes it isn’t. Say Jonah’s migration had already gone to production. Say the app was writing every cart edit to both the server and the local database. Turning the flag off would then leave old local carts on users’ phones. A later restart could sync those stale carts back over newer server data. Shoppers would see items they had already removed.

In that case, “I committed and moved on” is the wrong answer. The right one is a short, specific request. Jonah would ask for two more days to remove the double write and clear the local table. He would say that the payment date still holds. He would also say what users would hit if those two days were skipped. That still counts as disagree and commit. Jonah backs the new priority. He just won’t leave a known data bug behind it. In the interview, name the risk, the days you asked for and who approved them.

Leave a Comment