Microsoft As Appropriate Interview: Your Last Interviewer May Know Your Weakest Round

A Microsoft as appropriate interview, or AA, is the final interview of a loop. The name is old Microsoft slang. Getting one doesn’t mean you passed. Raymond Chen is a longtime Microsoft engineer. In a 2023 post, he explains the history. The final interview once happened only if the earlier feedback justified it. At some point that changed. Now it always takes place, whatever the other interviewers said. But Chen says the role itself didn’t change. In his account, the last interviewer still gets the earlier interviewers’ feedback. So the AA is a scored round. The person running it may know your weakest moment of the day. Prepare for it like this. Keep a one-line log after each earlier round. In the final round, close the one gap that matters. Do it once and keep it brief. This post follows one Android engineer through a Microsoft loop. It ends with the exact words he used in the last round.

The AA interview once meant the loop was still going well

Chen’s post on The Old New Thing, Microsoft’s developer blog, is dated October 2023. It describes how the name started. A typical loop had four or five interviews in one day. The recruiter opened and closed it. Every interviewer’s result went to the recruiter and to the final interviewer. That final person was usually more senior in the hiring group.

The final interviewer read the earlier recommendations first. If the candidate no longer looked worth pursuing, the day ended early. The last interview happened only when it was “appropriate.” So in that era, reaching the AA was real news. It meant the earlier rounds hadn’t sunk you.

Today the final interview happens whatever the feedback said

Chen writes that the AA later stopped being optional. It now “always takes place, regardless of the feedback from the other three interviewers,” in his words. The role also rotated. Every interviewer served as the AA for some candidates, so it was no longer one fixed senior person. The name faded too. Chen says the final interview no longer carries the AA label, though veteran interviewers still use it.

If every candidate gets a final interview, being scheduled for one tells you nothing. Your recruiter may not even call it the AA. They may just say you’ll meet one more person.

Chen says directly that one part didn’t change. The role of the final interviewer stayed the same. In his description, that role starts with reading what the earlier interviewers recommended.

In a Microsoft as appropriate interview, the last interviewer may know your weakest round

Chen describes how the role worked in his post. He doesn’t publish a current Microsoft policy. Teams may also run it differently. Still, the safe assumption follows from his account. Your last interviewer may walk in with notes from your morning. They can use the hour to test the area those notes flagged.

This is the opposite of a follow-up round scheduled after a debrief. Our post on an additional interview after the onsite says not to mention the earlier rounds. That interviewer may start from zero. The AA is different, because its interviewer may not start from zero. A gap you leave unexplained may be the first thing they probe.

It also differs from Amazon’s version. Our post on the Amazon bar raiser interview covers a reviewer from outside the team. That round is judged in a debrief. The AA, in Chen’s account, sat inside the hiring group.

Omar’s loop left one round with untested code

Take an Android engineer we’ll call Omar. He’s interviewing for a mid-level role. Microsoft’s How we hire page says most interviews have two to four conversations. Each lasts up to an hour. Omar’s loop has four, all on Microsoft Teams.

Round two is coding. Microsoft’s technical interviewing page calls it a short 45-minute round. The same page lists four areas it evaluates. They are problem solving, design, coding and testing. Testing covers boundaries, error conditions and corner cases in your own code.

Omar’s problem is to merge overlapping time ranges for a calendar view. His approach is right. He sorts by start time, then merges in one pass. But he spends too long on the merge loop. He runs one happy-path example with two minutes left. He never checks an empty list or two ranges that only touch. The interviewer asks, “What would you test next?” Omar names the empty list and runs out of time.

This is roughly the code he ends the round with.

data class Slot(val start: Int, val end: Int)

fun merge(slots: List<Slot>): List<Slot> {
    val result = mutableListOf<Slot>()
    for (s in slots.sortedBy { it.start }) {
        val last = result.lastOrNull()
        if (last != null && s.start < last.end) {
            result[result.size - 1] = Slot(last.start, maxOf(last.end, s.end))
        } else {
            result.add(s)
        }
    }
    return result
}

Rounds three and four go well. The fourth interviewer, a senior engineer, is the last of the day.

Write a one-line log in each break

Omar can’t know what the earlier interviewers wrote. He can know what he would want to fix. So in each break, he writes one line per round. The line names what he would change if he had five more minutes. Then he picks the single gap that matters most.

His log after round three looks like this.

R1 Android domain: fine. Could have named the lifecycle owner sooner.
R2 Coding: merge was right. Never tested empty list or touching ranges.
R3 Design: fine. Skipped the cache eviction question, interviewer moved on.

Round two is the clear pick. It’s the only gap in an area Microsoft says it scores. It’s also the easiest to close in under a minute. Omar prepares one short statement. It covers the cases he missed and what his code does with each. He checks the code in his head first. An empty list returns an empty list, because the loop never runs. Ranges that touch, like 9 to 10 and 10 to 11, depend on one comparison. His overlap check is s.start < last.end, and 10 is not less than 10. So the code keeps them as two ranges. For a calendar, he’d merge them, so the comparison should allow equal.

The other two lines stay in the log. Raising three things would turn the final round into a review of the morning.

Close the gap once, when the interviewer gives you room

Omar doesn’t open with it. The last interviewer has an agenda, so he answers their questions first. The round turns out to be a mix. There’s a short design question, then questions about his current team. With ten minutes left, the interviewer asks if there’s anything he wants to add. That’s the opening.

One thing, briefly. In my coding round this morning, I ran out of time before testing edge cases. I’ve thought about the two I missed. An empty list returns empty, because the loop never runs. Two ranges that touch, like nine to ten and ten to eleven, stay separate in my version. That’s because I compared with strictly less than. For a calendar, you’d want those merged. So I’d change that comparison to less than or equal, then add both cases as tests. That’s all on that. I’d like to ask you about the team’s release process.

If the interviewer had asked “How do you think today went?” Omar would use the same answer. If they never give an opening, he says nothing about it. The final round still counts on its own terms.

Why each part of Omar’s answer works

“One thing, briefly” tells the interviewer how long it will take. They can let him finish without watching the clock.

“I ran out of time before testing” names the gap as a fact. There’s no apology. Our post on bombing one round of an onsite explains why to skip the apology.

The two cases are concrete. One of them even changes the code. Saying “I’d test edge cases” is what every candidate says. Finding that touching ranges stay split shows the testing habit Microsoft lists. It also shows Omar checked his own code after the round ended.

“That’s all on that” closes the topic. Then he moves to his own question. The interviewer gets back the rest of the hour. The last thing they hear isn’t about the morning.

When to leave the earlier rounds alone

Skip the correction if your log has no real gap. Inventing one to look self-aware just points at a weakness nobody noted.

Skip it if the gap is only a feeling. “I think round three went badly” with no concrete fix gives the interviewer nothing new to write down.

Also skip it if the final round has a fixed format with no open time. Ask before the loop. Microsoft’s careers page says your recruiter will share a timeline for next steps. The same call is a fine time to ask who you’ll meet last and what the round covers. Match your prep to that answer. Keep the one-line log either way.

Leave a Comment