Earning trust: what interviewers score and how to choose the conflict story

This is the most-failed competency in the round. It is failed at selection more than at delivery, and it is failed because most candidates misunderstand what the question is for.

This lesson corrects that misunderstanding first, then sets out the signals — the negatives here are heavier than in any other competency — and the phrasings, two of which are traps. It ends with the selection tests and the places to look for a story you have probably not classified as conflict.

The definition

Read that last clause again. Candidates instinctively select the story where they turned out to be correct, because being correct feels like the win condition. It is close to irrelevant. A story where you were wrong, found that out, and handled it well scores higher than a story where you were right and the other person capitulated.

What it is not

It is not conflict resolution as a skill. Interviewers are not looking for mediation technique. They are looking for evidence about how you behave when someone blocks you.

It is not agreeableness. Avoiding disagreement scores as badly as handling it badly. "I try to find common ground and keep things positive" describes someone who will nod through a bad decision, which for a senior engineer is a real cost.

It is not being right. See above.

It is not conflict about people. A story about a difficult personality is a story about someone else. The competency is about disagreements over what to do.

What the interviewer is actually predicting

Three things, and they are all about the future rather than the past:

1. Will you tell me when I am wrong? A senior engineer who agrees with everything is worth less than one who does not.
2. Can I put you in a room with a difficult stakeholder? This is a question about whether you make situations worse.
3. When a decision goes against you, what do you do for the next six months? The candidate who relitigates is a permanent tax on a team.

Question 3 is the one interviewers care most about and the one candidates almost never address unprompted. Disagree and commit is entirely about it.

Signals scored

Those predictions are made from specific things you say. Four positives, five negatives. The negatives here are heavier than in any other competency — several of them end loops on their own.

The positive indicators

1. You sought their reasoning before arguing. The observable form is that you can state what they wanted and why, in their terms:

"I asked what would have to be true for them to be comfortable, and the answer wasn't anything I'd predicted."

2. You can state their position at its strongest. The single most diagnostic sentence in the competency:

"Their argument, and it was a good one, was that their team would inherit the operational load for a system they had no headcount to run. That was true and I hadn't accounted for it."

3. You separated the person from the problem. No characterisation. Not "they were being territorial" — "they had a constraint I didn't know about".

4. Something changed in you. Your design, your plan, your approach, or your behaviour. A story where the only change was the other person's mind reads as a debate you won.

The negative indicators

The one that decides it

Whether you can state the other side's argument well enough that they would recognise it.

Interviewers listen for this specifically because it cannot be faked in real time. If you did not understand their position at the time, you cannot reconstruct it now under questioning — you can only produce a version of it that is easy to defeat.

The two versions, from the same candidate about the same event:

Strawman: "They were worried about the extra work, but it was only going to be a couple of hours a week."

Steel version: "Their argument was that on-call for the new service would land on a four-person rota that was already at nine pages a week, and they'd had two people leave in a year partly over on-call load. The 'couple of hours a week' framing I was using measured the wrong thing — the cost wasn't hours, it was interruptions to a team that was already fragile."

The second candidate has demonstrated they listened. That is the whole competency.

Questions asked

Eight phrasings across four distinct relationships. Which relationship the question names changes what is being scored.

1. "Tell me about a conflict with a colleague."
2. "Tell me about a time you disagreed with your manager."
3. "Tell me about a time you had to push back on a stakeholder or a product manager."
4. "Tell me about a time you received feedback you disagreed with."
5. "Tell me about a time you had to work with someone difficult."
6. "Tell me about a time you had to deliver bad news to someone senior."
7. "Tell me about a time you were wrong about something and had to change your position."
8. "Tell me about a decision that went against you. What did you do next?"

What each is really probing

Question 5 is a trap

"Tell me about a time you had to work with someone difficult" invites you to describe a difficult person. Every word you spend characterising them is a negative mark.

Answer it as a question about a situation:

"The situation that was hardest for me was a review process with a staff engineer on another team that stalled for three weeks. I'd describe the difficulty as structural rather than personal — they had a constraint I didn't know about and I was pushing without understanding it. Here's what happened —"

That reframe takes eight seconds and changes what the interviewer writes down.

Question 4 is harder than it looks

"Feedback you disagreed with" is a double question. The interviewer wants to know both whether you can hold your ground and whether you can be wrong. The failure modes are symmetrical:

• Full capitulation — "I realised they were right and changed" — with no evidence you considered whether they were.
• Full rejection — "I explained why the feedback didn't apply" — with no evidence you looked for the part that did.

The strong answer usually contains a split: "About a third of it I thought was wrong and I said so. About two-thirds was correct and I hadn't seen it. Here's the two-thirds."

Building the story

The selection trap here is worse than in any other competency, and it accounts for most of the failures.

The trap

Candidates search their memory for "conflict" and find one of two things:

Nothing. Because they are picturing shouting. Ninety percent of engineering conflict is a design review that went four rounds, a stalled approval, or a quiet disagreement about priority that nobody named.

Something trivial. Tabs versus spaces. Which library to use for date parsing. A story with nothing at stake produces nothing to score, and it also produces a second, unspoken mark: this person's biggest disagreement in five years was about formatting.

The three tests

1. Real disagreement. Two people wanted materially different things, and both positions were defensible. If one side was clearly wrong, this is a correction, not a conflict.

2. Real stakes. Something was at risk: a date, a design that would be hard to reverse, money, someone's workload, a user outcome. "We would have had slightly messier code" is not stakes.

3. You changed something about yourself. Your design, your plan, or your behaviour. This is the test that eliminates most otherwise-good stories, and it is non-negotiable — a story where you changed nothing is a story about persuading someone, which is a different competency.

Where to look

Go through the last two years for these, in rough order of yield:

Level calibration

The same shape of story reads at different levels depending on who you disagreed with and what was at stake.