Mock Interviews: Why They Work and How to Run One

A candidate can solve hundreds of practice problems and still go quiet a few minutes into a real interview. They knew the material. What they had never done was perform it: aloud, watched, against a clock, with a stranger deciding the outcome. Mock interviews exist to close that gap, and they are the part of preparation people most often skip, because a mock is uncomfortable in a way that another evening of solo study is not.
A mock only helps if it is run properly. A friendly chat over a problem, followed by "that seemed fine", teaches very little. This guide covers what a good session looks like minute by minute, what the feedback should contain, how many mocks different situations call for, and how a peer, an AI interviewer and a paid coach compare, including what a coach costs.
What rehearsal trains that study does not
Studying alone builds knowledge. An interview tests whether you can get at that knowledge while doing three other things.
- Thinking and talking at once. Explaining an approach while writing code or drawing a design is a divided-attention task, and it has to be trained as one. It is why capable engineers fall silent for five minutes and lose the interviewer.
- Seeing yourself. Alone, you cannot notice your own habits: the "we" that hides what you did, the story whose setup runs for three minutes, the complexity claim you never justified. An observer catches in one session what you would not find in a month.
- Being watched. The first time someone watches you work a problem, a good part of your attention goes to the fact of being watched. After several sessions, most of it is back on the problem. Only being watched produces that change.
- Being interrupted. A real interviewer asks "why not a relational database?" in the middle of your sentence. Handling that calmly, and returning to where you were, is a skill with no solo version.
The conditions that make a mock count
- Full length and no pauses. If the real round is 45 minutes, the mock is 45 minutes. No stopping to look something up, no restart after a bad opening.
- The real tools. Camera on, the same kind of shared editor or whiteboard the company uses. If the round is in a plain editor with no autocomplete, practise in one. If a round will be in an office, do one mock standing at a whiteboard; the guide to in-person interviews explains what changes.
- The right round. Coding, system design, behavioural, or a company-specific format. A mock for the wrong round is pleasant and wasted.
- An interviewer who knows the bar. A friend reading a question aloud is better than nothing. Someone who has run real loops, and knows what a passing answer at your level sounds like, will ask the follow-ups that a real interviewer asks.
- A debrief straight afterwards, and notes in writing. Memory of what happened fades within the hour, on both sides.
A 60-minute mock with a coach, minute by minute
This is an example of how a one-hour session can be divided, written for this guide. Coaches run their sessions in their own way, and a system design or behavioural mock shifts the middle section, so use it as a picture of where the time should go.
| Minutes | What happens |
|---|---|
| 0 to 5 | Setup: target company and level, the round being rehearsed, your interview date, anything you want watched |
| 5 to 10 | The interviewer goes into role and opens as a real one would, with a short introduction and the first question |
| 10 to 45 | The round itself, uninterrupted except for the follow-ups and hints a real interviewer would give |
| 45 to 47 | You get the last two minutes a real round gives you for your own questions |
| 47 to 57 | Debrief: the verdict first, then the two or three moments that decided it, replayed in order |
| 57 to 60 | The action list: what to drill before the next session, and what the next mock should test |
Three things are worth noticing in that division.
The setup is short on purpose. Send the context beforehand (the job description, the round, your level, the date) so that the hour is not spent on introductions.
The round is not interrupted for teaching. If the interviewer stops at minute 20 to explain a better approach, you have had a tutorial. Tutorials are useful, but you can no longer learn how the round would have ended. Notes are saved for the debrief.
The debrief starts with the verdict. "At the senior bar this would have been a no-hire, mainly because of the first ten minutes" is hard to hear and tells you more than twenty minutes of balanced comments. Ask for it in that form if it is not offered.
What written feedback should contain
Ask for the feedback in writing, during the session or within a day. Good written feedback has five parts.
- A verdict at a named level. Hire, no hire or borderline, for a stated level and round. "Strong for mid-level, short of senior" is usable. "Pretty good" is not.
- Evidence with timestamps. What you said or did, and roughly when. "Minute 14: started coding before stating the approach; the interviewer could not follow until minute 22."
- How it reads from the other side. The same behaviour translated into what a real interviewer would write down. Silence reads as being stuck. A long setup reads as poor prioritisation. This part is what you cannot get from a recording of yourself.
- Two or three fixes, each one a drill. "State the brute-force approach and its cost before writing any code" can be practised tomorrow. "Communicate better" cannot.
- What to test next time. The next mock should check whether the fixes held, on a different question.
Keep every feedback note in one file. After three or four sessions you can read down them and see whether the same problem keeps appearing. When the notes start naming new, smaller problems each time, you are close to ready.
How many mocks, by situation
These are working rules, not measured figures, and the right count depends on how the first one goes.
| Your situation | A sensible number | Why |
|---|---|---|
| First loop in several years | 4 to 6 across round types | Everything is rusty at once, including being watched |
| Interviewing now, one round keeps failing | 2 to 3, all on that round | Find the cause, fix it, confirm the fix |
| Staff, principal or manager loop | 3 to 4, mostly deep dive and stories | The level is decided in the rounds you cannot rehearse alone |
| New graduate or career changer | 5 or more, many with peers | Volume matters most, and peers at your level are easy to find |
| One specific company in under two weeks | 2: one now, one three days before | One to diagnose, one to rehearse |
| Just rejected after an onsite | 1, on the round you think went wrong | Check your own diagnosis before preparing again |
Whatever the number, the first mock should come early. Most people guess wrong about their weakest round, and a diagnostic session in the first week redirects the rest of the preparation. A mock the night before the interview is too late to change anything and mainly affects your sleep.
Leave three or four days between mocks on the same round. The gap is when the drills happen. Two mocks on consecutive days measure the same weakness twice.
Peer, AI and coach compared
The three do different jobs, and most preparation plans use more than one.
| Type | What it is good for | What it cannot do | Cost |
|---|---|---|---|
| Peer | Volume, being watched, practising talking while solving | Tell you whether the answer passes at your level | Free, paid in favours |
| AI interviewer | Repetition at any hour, saying stories aloud, structure drills | Read the room as a person does, or judge level reliably | Often free |
| Coach | Calibration against a real bar, realistic follow-ups, a verdict | Volume, unless your budget is large | Paid per session |
Peers are the way to get repetitions. Trade sessions with someone preparing for a similar level, take turns as interviewer, and use a question neither of you has seen. Playing the interviewer is useful in its own right, because you see how little of a candidate's thinking is visible from the other side. The limit is calibration: two people who have never sat in a debrief can agree an answer was good when it would not have passed.
AI interviewers have become a reasonable way to practise speaking. On techinterview.coach the practice interview is a voice interview with an AI interviewer on the round you choose (behavioural, system design, coding talk-through or salary negotiation), optionally based on a job ad you paste in, with a written recap of what worked and what to drill. It is free; sign in to start. Use it for the tenth telling of a story or a design opening late at night. Do not treat its recap as a hiring decision.
A coach is for judgment. Someone who has made hiring decisions can tell you whether an answer would pass at your target level, which is the one thing the other two cannot supply. That makes the timing matter more than the count: one session early, to find out where you stand, and one near the end as a dress rehearsal. The mock interview coaches on techinterview.coach are listed with their backgrounds, so you can choose someone who has interviewed for your kind of role.
What a mock with a coach costs
Prices are set by each coach and they change, so check the page on the day. Measured on techinterview.coach in November 2026:
- 34 of the 40 coaches offer a mock interview, in 39 listings between them, because a few coaches list more than one.
- 35 of the 39 are a single 60 minute session. Those single sessions cost between $79 and $458 for a first booking, and 25 of the 35 are priced from $149 to $179.
- The other four are two longer sessions (75 and 90 minutes) and two packs of two or four sessions.
- 36 coaches also offer a shorter intro call, from $21, if you want to check the fit before paying for a full session.
Whether that is worth it depends on what is at stake and what you can get free. If you have a colleague who interviews regularly and will be honest with you, start there. The guides to what interview coaching costs and whether it is worth it go through the calculation.
Getting full value as the candidate
- Brief the interviewer. Send the company, level, round, date and the one thing you most want watched. "Tell me if I talk too long before coding" gets you a sharper debrief.
- Treat it as real. No restarts, no "can I look that up".
- Struggle before you take a hint. Ask for the hint a real interviewer would give, at the moment they would give it.
- Ask for the harsh version. Most people soften feedback unless asked not to. Say you want the debrief they would give a hiring committee.
- Repeat the format, not the question. Do the same type of round on a new question a week later. Solving the same question again shows only that you remember it.
- Record it if the interviewer agrees. Ten minutes of watching yourself shows the filler words and the long silences faster than any written note.
Where coaching fits
The hardest ingredient to find in a mock is an interviewer who knows the bar, and for most people that means paying for one or two sessions and filling the rest with peers and AI practice. The mock interview page lists the 34 coaches who offer one, with their backgrounds and prices, so you can book an early diagnostic session and a rehearsal close to your interview date. If you are not ready to be watched by a person yet, start with the practice interview, which is free once you sign in, and book the coach when the recap stops finding large problems.
Related guides
- A 4-week prep plan for senior engineers: where the mocks sit in a full schedule.
- Interview coaching cost: what coaches charge and what drives the price.
- Is interview coaching worth it: the calculation for and against paying.
- STAR method: the shape for the stories a behavioural mock will test.
- System design interview questions by level: prompts to hand to a peer for a design mock.
- AI in coding interviews: what to ask the recruiter, so the mock matches the real round's rules.