The Difference Between Practising Answers and Practising Under Pressure

He had his whole career ready, and lost it on one question

The candidate was prepared. He walked through his career clearly, company by company, in order, with no gaps and no rambling. If preparation was the thing being tested, he passed.

Then I asked why he decided to leave each of them.

That's where it fell over. Not because the answer was damaging, and not because he had something to hide. He had prepared for exact questions, and this one sat a level below the questions he had prepared for. He knew the story of his career. He hadn't thought about the decisions inside it.

None of that was a knowledge problem. He had the material and he had rehearsed the material. He just couldn't go anywhere the rehearsal hadn't already been.

The distance between having answers and being able to keep talking once the answers run out is the part most interview prep never touches. You can write out every question on every list and still lose the interview on the second question in a sequence.

Rehearsed is not the same as bad

I want to be clear about what I'm not saying, because "don't memorise your answers" is advice I disagree with.

I can usually tell when someone is running a prepared answer, and the tell is timing. They start talking the instant my question stops, with no pause at all, and the answer runs long. There's no moment anywhere in there where they consider what I actually asked.

Usually it's fine. The answer does address my question, and it's often better than the improvised version would have been: better organised, better chosen. Someone in the middle of a job search has done a lot of interviews that month, so arriving prepared is the natural result. I don't hold it against anyone. I listen.

So rehearsal isn't the failure. A prepared answer is a surface, and the next thing that happens in the room is a test of how deep it goes.

The follow-up is where the interview actually happens

My first question is a prompt. The second one is the interview.

Here's the kind of follow-up I mean. Somewhere in a prepared answer, a candidate mentions in passing that a team they worked on was the best team they'd ever worked with. Nice line. It's in the script. So I ask:

You said it was the best team you ever worked with. Could you explain what the criteria are, why this team is the best for you?

I'm not trying to catch anybody out. I want to know what they value in a team, and this is the fastest route to it. But look at what it does to a prepared answer. The question is built out of their own throwaway line, so there was no way to script it. I made it out of a sentence they'd said a moment earlier.

What happens next is visible from across the table. People stop reciting. You can watch them start thinking, genuinely looking for an answer instead of retrieving one. That's usually the first unrehearsed thing I've heard from them.

This is also the part that doesn't get better by adding more prepared answers. There's no list of the top fifty follow-up questions, because a follow-up isn't a question from a bank. It's a question about you, assembled live, out of something you just said. The only thing that prepares you for it is having practised thinking out loud while somebody watches.

The final round I lost to a SQL question I knew

I've been on the wrong end of this too.

I'd passed several stages at one company and reached the final interview. I prepared for a culture conversation, because that's what I'd decided it was going to be. The interviewer turned out to be a technical guy, and he asked me technical questions.

They were questions about SQL, something to do with sorting, and they were easy.

I went blank.

I knew that material the way you know things you've used for years, which is exactly why I hadn't looked at it. And it came back to me right after the interview ended, walking away from it, when nobody was asking. I didn't pass that stage.

What I took from it wasn't "revise SQL". It was that I had prepared for a format instead of preparing for an interview. I decided in advance what kind of conversation it would be, and everything I did was tuned to that decision. Now I try to be prepared for any kind of question, which in practice means being ready to think in front of someone rather than to retrieve.

Nerves get about fifteen minutes

Nervousness has a specific sound from my side of the call: long pauses, unsure speech, filler words, and a general sense that the person is managing themselves rather than talking to me.

None of it counts against you at the start, because I expect it. Most people arrive tense and settle within five to fifteen minutes, and once they've settled it doesn't affect the result at all.

The people it costs are the ones who never settle. If someone is still in that state at the end, it does show up in the outcome, and not because I'm scoring their delivery. It's that the hour finished and I still hadn't seen them think. There's nothing to put in the debrief except how they sounded.

You can watch the same effect stretched across a whole loop. People get more relaxed stage by stage. By the final round they're more prepared for that specific company, they know the shape of the questions, and they're simply easier to talk to than they were in the first conversation.

Which says something uncomfortable about the first interview in any search. It gets the worst version of you: least prepared, most stressed. That's normal and it's fine. It's also a good reason to make sure the first stressful conversation you have about your own work isn't the one with the company you actually want.

The hour you spend on another algorithm is usually the wrong hour

The standard advice for engineers is to focus on hard skills: more algorithms, more system design, more of whatever can be measured. That's the advice I think most often makes people worse in my loops.

Hard skills are the improvable part. They always were, and with AI sitting in everyone's editor they improve faster now than they used to. How you explain your work is the slow one.

I'm not arguing that technical depth stopped mattering. I'm arguing about where the marginal hour goes. If your answers are already written down and you have one hour left, another algorithm isn't the thing standing between you and the offer.

What pressure practice actually looks like, and when to stop

Start with the cheapest possible version of it: say your answers out loud. Not necessarily to an AI tool. Out loud, to a wall, in the car, on a walk. Written answers you've never spoken are untested. They're notes about an answer rather than an answer.

An answer that reads well on the page is a different object once it's coming out of your mouth at speed, and you want to find that out on a walk rather than in the interview.

After that there are two different exercises, and it's worth knowing which one you're doing. I split them in the product from the first version, because they aren't the same activity.

Practice is the thinking room. One skill at a time, at your own pace, with room to stop and start and take as long as you need to find the words. Nothing is timing you.

A mock interview is the pressure room. A random voice, a clock, and a flow of questions rather than a list of them. It isn't meant to be gentler than a real interview, or harsher. What matters is that you don't set the pace.

Use the first to find the words. Use the second to find out whether they survive being interrupted.

There's a ceiling to this, and you can feel it when you reach it. You already know all the answers, you've stopped thinking, and you start to sound robotic. Repetition past that point makes you worse, because robotic is exactly what I notice from the other chair.

When you hit it, stop. Or change the drill and practise follow-ups instead: take your own answer, find the line in it you said without thinking, and explain that. It's the part of an interview you can't memorise, which makes it the only part still worth the hour.