How to Run a Mock Interview With a Friend: Script, Timing, and Rubric

Week 3 of the 4-week prep plan carries one instruction that gets skipped more than any other: run at least two full mock interviews under real conditions. When I ask people why they skipped it, the answer is usually some combination of money (the paid options aren't cheap) and logistics (asking a friend feels awkward, and neither of you is quite sure what "running a mock" actually involves). This article solves the second problem completely, and honestly it takes a big bite out of the first one too, because a friend who runs this kit well covers most of what paid tools charge for.

Everything below is free to use. The rubric in particular is the piece I can stand behind personally: it's a condensed version of what I graded real candidates against over a lot of interview loops.

Why real conditions expose what studying can't

I spend a fair amount of time woodworking, and the step that rescues every project is the dry fit. Before the glue comes out, you clamp the whole piece together, and every gap, every joint cut a hair proud, every panel that won't quite seat reveals itself while it's still cheap to correct. Glue-up is the real interview, and a problem you first meet during glue-up is a problem you meet in the worst possible way.

Solo studying is checking each joint on its own and never assembling the piece. Every component can measure perfectly and the whole thing still won't close square, because the failures live in the assembly: nerves, the clock, narrating while you think, a person reacting to what you say. You cannot observe the assembly without assembling it. In my experience a single full-dress mock surfaces more fixable problems than another week of grinding problems alone, and I've watched genuinely strong engineers lock up in their first one. If that turns out to be you, don't be too hard on yourself. Finding the gap in the dry fit instead of the glue-up is the entire reason you ran it.

Assign the roles and defend them

Friend mocks mostly fail through role collapse. Five minutes in, the interviewer starts helping, the candidate starts editorializing about what they'd normally do, and the session dissolves into pair programming with commentary. So agree before you start: for the whole block, your friend is the interviewer. They own the clock, they hold the hints, they don't teach, and all feedback waits for the debrief. If you're both prepping, run two blocks and trade chairs (the interviewer seat is a valuable rep in its own right, because you find out what silence feels like from the other side of the table).

The interviewer also chooses the problem, keeps it secret, and works through it themselves ahead of time. You can't guide someone through a maze you haven't walked.

Timing for each round type

• Coding, 45 minutes. Around 3 minutes of intro, 35 on one medium problem the candidate hasn't seen, 5 for the candidate's questions, and a small buffer.
• Behavioral, 35-40 minutes. Three stories at 8-10 minutes apiece including follow-ups, plus intro and closing. Draw prompts from a standard story bank: conflict, failure, ambiguity, impact.
• System design, 45-50 minutes. Follow the standard skeleton: about 5 minutes on requirements, 3 on estimates, 10 on high-level design, 15-20 on the deep dive, and 5 on tradeoffs and failure modes. The interviewer mostly steers ("walk me through how the feed gets generated") and protects the clock.

Keep a timer visible and stop on time, even mid-sentence. Letting the round run long feels kind and quietly wrecks the rep, because time pressure is a large part of what you're rehearsing.

The interviewer script

The opening, which your friend can read word for word:

"Thanks for making time. This is a 45 minute round. We'll spend most of it on one problem, and I'd like you to think out loud as you work. I'll ask questions along the way, and we'll save the last five minutes for anything you want to ask me. Sound good?"

The hint rules, which are what untrained interviewers get wrong most:

1. If the candidate goes silent for about a minute without announcing it, ask "can you tell me what you're thinking?" Real interviewers do this, and it also trains the habit of announcing pauses (the value of which I've written about in talking while coding).
2. If they're stuck on the same point for 4-5 minutes, give the smallest hint that could possibly unblock them, escalating only as needed. "What operation is expensive here?" comes before "is there a structure with faster lookups?" which comes before "consider a hashmap keyed on the character."
3. If they're on a wrong path they haven't noticed, ask one question and let them discover it. "What's the complexity of this approach?" or "can you trace it on this input?" usually gets them there without handing over the answer.
4. Don't teach mid-round, and hold back the cheerleading after every line. Real rounds run quieter than friends naturally want to be, and reading a neutral face is one of the skills under rehearsal.

For behavioral rounds, probe every story at least twice: "What was your specific contribution?", "How did the other person respond?", "What would you do differently now?", "What was the actual result, do you have a number?" Interrupt one story midstream on purpose and redirect it. Real interviewers probe in directions you didn't plan for, and a story that only works as an uninterrupted monologue needs more work.

The four-axis rubric

Score each axis 1 to 4, and have the interviewer write a sentence or two per axis during the round. Memory after a mock retains roughly the last five minutes plus whatever went worst, so live notes decide whether the debrief has anything real in it.

• Communication. Could a listener who can't see the screen follow the approach? A 1 is long unannounced silences with reasoning that only exists in the finished code. A 4 is a stated plan before typing, announced pauses, and decisions explained as they happen.
• Problem solving. A 1 charges into the first idea. A 4 names the brute force, calls out the pattern, and defends the chosen approach against at least one alternative.
• Correctness. Does it work? A 4 handles or at least names the edge cases and states complexity unprompted. In behavioral rounds this axis becomes: specific story, your actions at the center, and an answer to the question that was actually asked.
• Recovery. Scored only on what follows the first wrong turn or hint. A 1 ignores the hint or spirals. A 4 absorbs it, folds it in, and keeps structure. A flawless round leaves this axis blank, which is fine; in a real loop the interviewer escalates difficulty until it appears.

The debrief, in order

1. Candidate self-assessment, two minutes, before any feedback. Calibration is a signal of its own, and being wrong about how it went (in either direction, and I saw both regularly) is worth surfacing first.
2. The four axes, from the live notes, one concrete moment each. "At minute 12 you went quiet for two minutes and I lost the thread" lands far better than "communication could be better."
3. One thing to keep, two things to fix. Cap it at two, because a list of seven fixes rarely produces any fixes before the next round.
4. The closing question, spoken out loud: hire or no hire, and the single biggest reason. It's uncomfortable, and it's the most valuable sentence of the whole session, because it's precisely the sentence a real interviewer writes about you afterward. A fake round with a friend is the cheapest place you will ever get to hear it.

What a friend mock leaves out

Run seriously, this kit covers the large majority of what paid mock services provide. The honest remainder is three things: stranger pressure (a judge whose opinion of you isn't already settled raises the stakes in a way a friend can't), calibration (your friend has graded one candidate in their life, and it's you), and availability (friends have jobs, and a real prep push wants multiple reps in a week). Free peer mocks with strangers exist on Pramp (now hosted on Exponent, with 5 free credits a month), and interviewing.io offers paid sessions with professional interviewers at roughly $180-340 each.

If your schedule or your bench of willing friends runs thin, on-demand practice against something that responds, probes, and grades you is exactly what we built Preppable to do. But I'd genuinely rather you run this kit with one honest engineer friend than skip real-conditions reps entirely, because the dry fit only helps if you actually clamp the piece together before the glue comes out.