From applications to preparation: why we built Prept's Mock room

Why do candidates get so little feedback after an interview?

When we first started building Prept, we kept hearing the same frustration from candidates: there was no feedback they could act on. You go through an interview, wait for news, and get either a generic rejection or silence. Nobody tells you what you did well or which answer cost you the offer.

Without that, it is almost impossible to improve. So we set out to solve that problem first, with the interview debrief: a written review of a real interview once the call ends.

Why did we build a resume tool and a LinkedIn tool next?

Once users started getting feedback on their interviews, they came back with a new problem. In their words: "It's great that you can analyze my interview performance, but I'm not even getting enough interview calls. Something is wrong with my application."

They were right. The market was full of tools that scraped job descriptions and repeated their keywords line by line. Those applications all looked alike and sank into a pile of identical resumes.

So we built the resume optimizer to show your actual experience and how your background fits the role, instead of repeating the job description word for word. Every claim traces back to a fact you confirmed, which is what we call a fact-locked resume.

Then users hit the next bottleneck: standing out on professional networks. We built the LinkedIn optimizer to line a profile up with the new resume, so a candidate tells one story everywhere. For many of our users, that is when the first interview calls came in.

Why is learning during a live interview so costly?

With more interviews came a new gap. An early tester put it plainly: "It's all good, but isn't it too late for me to know what I'm doing wrong? By that time, I have lost an opportunity for a job that I really loved. How can you help me with that?"

That one piece of feedback reshaped our product plan. A debrief after a real interview teaches you on an opportunity you have already spent. You shouldn't have to lose an offer at a company you love to find out your answers needed work.

What is the Mock room?

The Mock room is a Practice tab inside Prept. You answer interview questions out loud, against a clock, and get a score. The score comes with how it was made up, the one change that would raise it most, and your own answer said better. Then you try again.

What sets it apart from a generic mock interview is that every answer is checked against your confirmed resume facts. If your resume says 22 engineers and you say "about 25 engineers", the Mock room catches it before the interviewer does. That kind of drift between the resume and the answer is what we wrote about in did my story change.

How does a practice answer work?

Each answer runs like a short slice of a real interview: a question, a clock, a cut-off and a follow-up. Press the space bar to start and stop an answer.

1. Practice home lists your questions, each with its last score.
2. A microphone check.
3. The question, with 10 seconds to think. With notes, short notes built from your facts are shown and the clock waits until you are ready. Without notes, the clock starts by itself at zero, as it would in a real interview.
4. Your answer, with a countdown of the time left. A "Bring it to the result" cue comes 15 seconds before the target (10 seconds on short questions). At the target it says "Time. Finish your sentence." After a grace period of a quarter of the target, at least 15 seconds, the recording stops, as an interviewer would cut in.
5. A follow-up question, the kind an interviewer would ask next. For example: "How do you know it was your work that moved it?"
6. Scoring, then the result.

Extra time in Settings makes every target 1.5 times longer. If you can't use a microphone, you can type an answer instead; it is scored without length and delivery.

How is your answer scored?

Each answer gets a score out of 10, built from six parts. Nearly half the score, 45%, is checked by code, so it comes out the same every time. The model judges the rest against written descriptions of each score band, with settings that give the same score for the same answer.

• Three levels: Not yet (below 5), Getting there (5 to 6.9) and Ready (7 and above).
• The one thing: the single change that would raise your score most, written as a short instruction to you.
• Say it like this: your own answer reordered and tightened, using only your words and your confirmed figures. If the rewrite adds anything you didn't say, it isn't shown at all.
• That's not what I said: if the transcript misheard a number, you can drop it and the score is recalculated.
• Your transcript marks the figures that match your resume, flags the ones that don't, and underlines filler words.
• Progress: you see the change since your last answer to the same question, and whether you managed it without notes. Each question keeps every answer you have given it.

What happens in your first free session?

Your first session has three questions, picked from your resume by a fixed rule rather than by the model. It takes about five minutes, and each answer is scored in the background while you take the next question. You can stop after any answer.

1. "Tell me a bit about yourself." (60 seconds)
2. "Walk me through [your biggest change figure] at [current employer]. How did you get there?" (90 seconds). It uses a figure that records a change, such as "the 18% increase in completed bookings", because that invites "how?".
3. "Walk me through your time as [earlier role] at [employer]. What did you own?" (60 seconds). With no earlier role, it asks "Why are you looking to move?" (30 seconds).

At the end you get a baseline that shows where you are starting. Without a confirmed resume, the session runs on general questions.

Which resume facts have you never said out loud?

Every confirmed fact on your resume gets a second status: said well, said differently, said but not yet well, or never said. Your latest answer that touched a fact decides its status. Each fact you have never said has a Practice this button, which makes a question about that fact.

After the first session there is always one question you haven't tried: the likeliest kind you haven't practiced yet, taken from the interview question bank. Each application also gets practice for its next round, shown above everything else.

• Before the recruiter call: the 5 questions the recruiter call prep wrote answers for, asked about that employer.
• After it: the likeliest 4 questions for the next round, one of each kind.
• A reminder the day before: enter the date of your next interview on an application and Prept emails you from 08:00 UTC the day before. It names the round and your answer furthest from Ready, or tells you that you haven't practiced yet.

Why are we adding one question a week?

Instead of dumping thousands of generic prompts into the app at once, we are taking a deliberate approach: over the next three months, we are adding one new question every week. Raw quantity tends to lead to shallow feedback, and we would rather every question earn its place.

Every question we add is tested first: we check the prompt, refine the scoring criteria and verify that the feedback is accurate and something you can act on. One question a week keeps the feedback to that standard, and gives you a focused target to practice each week without feeling overwhelmed.

What does each plan include?

Practice has its own minutes, separate from your interview recording minutes, so practicing never uses up a real debrief. The app shows them as answers left: an answer with its follow-up takes about one and a half minutes. You are only charged for answers that get scored.

• Free: the first session, which never counts, then 10 practice minutes once, including one question from your own resume.
• Free week (trial): 60 practice minutes a week and every feature.
• Search: 240 practice minutes a month, with bank questions and round practice. See pricing for the current price.

Silence, answers under 15 words, failed transcriptions, Start over and answers cut off by a server restart cost nothing.

What happens to your voice recording?

Your voice is deleted as soon as it is transcribed. The score, the rewrite and the resume check all work from the words. Answers that end any other way keep no audio, and an hourly sweep deletes anything a crash left behind.

The Mock room never listens outside an answer, never shows a live transcript while you speak, and stops recording if you leave the tab. An answer stays on your device until the server confirms it has it, so a dropped connection or a closed tab doesn't lose it. Google's Gemini transcribes word for word, keeping the filler words, and scores the parts the model judges.

How do you try the Mock room?

The Mock room is live in Prept now. Log in to Prept and open Practice, answer this week's question, and read your debrief. Practice like this happens before the interview, which is the side of the line we wrote about in AI interview help that isn't cheating.

We build Prept from how you use it, so try the new feature, tell us what you think, and happy job hunting.