In a design review, a respected senior engineer proposes an approach you believe is wrong. The room defers to them. How do you handle it?

What the interviewer is testing

Whether you can be effective without authority, which is most of the architect's job. It is also testing whether you assume you are right.

Before disagreeing

Check whether you are. They may have context you lack — a prior incident, a constraint from another team, a commitment already made. Ask a genuine question first: "What made you rule out X?" The answer either gives you the missing context or reveals that it was not considered. Both outcomes are progress, and the question costs nothing.

The disagreement itself

Separate the claim from the person. Disagree with the design, in public, specifically, and without preamble about respect that signals the opposite.

Argue from consequences, not preferences. Not "I would do it differently" but "with this approach, when the payment provider is slow, every checkout thread blocks — here is the failure path". Consequences are checkable; preferences are a status contest.

Bring the alternative fully formed. Criticism without a worked alternative reads as obstruction, and the room is right to discount it.

Name what would change your mind. "If the provider's p99 is under 100 ms, my concern disappears — do we know?" This converts a disagreement into a question with an answer, and it lets the other person concede without conceding.

If the room still disagrees

Distinguish reversible from irreversible. If it is a two-way door, let it go — being right about a library choice is not worth the capital, and you can revisit it cheaply. Spend your credibility on the decisions that are expensive to undo: data models, service boundaries, public contracts, security posture.

If it is irreversible and you are still concerned, escalate transparently. Not behind anyone's back: "I do not think we agree, and this one is hard to undo. Can we get a second opinion from X?" Said in the room, this is professional. Said afterwards in a private message to a manager, it is not, and it will cost you more than the decision was worth.

Then commit properly. Disagree and commit is only credible if the commit is real. Record the concern in the ADR — including your dissent and its rationale — and then help make the chosen approach work. The ADR is what lets the decision be revisited on evidence later without anyone having to relitigate the meeting.

What a strong answer adds

Recognising that being right is roughly half the job and being heard is the other half — and that the architect who is right and abrasive gets routed around, after which they are not right about anything because they are no longer in the room.