01Mock interview · Interview preparation, scored

You have read your answers. You have never said them out loud.

Answer three questions out loud, free, and read what an interviewer actually heard: every answer scored, and your weakest one rewritten in your own words. The pass takes it to twenty questions.

Start my free mock See a real report

No card needed. Three questions, about six minutes.

First mock free Scored on relevance, specificity and structure Your words, never a model answer Money back on the report

OUT OF 100

Your lowest

58/100

Q4, system design, no trade-off stated

Structure

4 of 5

answers landed setup and approach

Say it this way

Rewritten from your own words, not a model answer

38out of 100

A real answer from a four-year engineer to "walk me through a system you designed." Microservices, Kafka, it was scalable. No number, no trade-off, no decision. He had designed it. He had never explained it to anyone who was not already on the team.

WHAT WE HEARD

"So we moved to microservices, and we used Kafka for the events, and it was basically much more scalable than before. There were some issues initially but we sorted them out. It worked well."

Setup ✗ · Approach ✓ · Trade-off ✗ · Outcome ✗

02See a real report

This is the report Aarav got. Yours takes ten minutes.

Five questions, scored on role fit, specificity and structure, with his weakest answer rewritten from what he actually said. Read all of it.

app.clearround.ai/report/mock-aaravSAMPLE

Aarav Mehta · SDE-2, four years, interviewing at Razorpay. A real report the product produced, not yours.

Reading the full sample

Mock interview report

5 of 5 answeredReady

VERDICTREADY

/ 100

Strong 4 · Adequate 0 · Needs work 1

Best on “Can you walk me through the architecture of the settlement service you own, focusing on how you ensure data consistency and handle failures?”. Lowest on “What eviction policy did you choose for the Redis cache of refundable balances, and why?”.

The answer needed the selected Redis eviction policy, the cache's source of truth, and the reasoning about correctness, expiry, memory pressure, and cache misses.

Role fit75Specificity77Structure73

→NEXT MOVEAnswer the exact Redis question first, then explain whether the cache is rebuildable and how your policy prevents eviction from causing incorrect refundable-balance decisions.

01 / 04 Verdict

◆

Readiness Summary

Readiness score, the reasons behind it, and the answers to fix. Everything else in this report supports this page.

✓

Reasons to relax

Strengths

What already lands, and where you can lean in.

4strong

Strong answers · 4

01Can you walk me through the architecture of the settlement service you own, focusing on how you ensure data consistency and handle failures?93

02Describe a time you introduced a new caching strategy to improve payment processing latency and how you measured its impact.92

03Can you explain how you handle duplicate or out-of-order settlement events to maintain accurate seller balances?82

04Tell me about yourself and your recent work at Swiftcart Technologies.77

Fix first

Weak answers

What loses the round if ignored, and the gap to close first.

1to fix

Lowest scoring · 1

01What eviction policy did you choose for the Redis cache of refundable balances, and why? — The answer needed the selected Redis eviction policy, the cache's source of truth, and the reasoning about correctness, expiry, memory pressure, and cache misses.28

02 / 04 Answers

▥

Answer by answer

5 questions, scored on relevance, specificity and structure. Open any to read the transcript beside the feedback.

Q1Tell me about yourself and your recent work at Swiftcart Technologies.

3srelevance · specificity · structureHide

What we heard

I am a backend engineer with four years on payments and orders at Swiftcart. Right now I own the settlement service that pays around forty thousand sellers every week. Before that I built the order cancellation flow and moved our notifications onto Kafka. I am on call one week in four, and I mentor two junior engineers. I am looking at Razorpay because I want to work on money movement at a scale where correctness is the whole job, not a feature.

Relevance 96Specificity 88Setup ✗ · Approach ✓ · Outcome ✗

✓ You made your experience concrete and relevant by naming payments, orders, settlement, Kafka, on-call ownership, and mentoring responsibilities.

✗ You did not state a specific situation or problem you faced, or explain what resulted from the work you described.

Why you scored 77

The answer is highly relevant and substantially specific, but it is primarily a concise inventory of responsibilities rather than a developed example. Setup is false because no concrete situation or problem is stated, and close is false because the Razorpay motivation is an intention rather than an outcome produced by the work.

How this could sound

I am a backend engineer with four years of experience in payments and order systems at Swiftcart Technologies, and I currently own the settlement service that pays around forty thousand sellers every week. My recent work has included building the order cancellation flow and moving notifications onto Kafka, with my responsibilities spanning on-call one week in four and mentoring two junior engineers. A strong example I would walk through is → the specific settlement, cancellation, or notification problem you solved, including the decision you made and why. I would close with → the observable result or business impact you can defend, such as the relevant reliability, correctness, or operational measure. I am interested in Razorpay because I want to work on money movement at a scale where correctness is the whole job, not a feature.

Q2Can you walk me through the architecture of the settlement service you own, focusing on how you ensure data consistency and handle failures?

3srelevance · specificity · structureFeedback + model answer

What we heard

The biggest one was moving settlement from a nightly batch to event-driven. The batch ran at 2 am and if it failed, sellers waited a full day. I designed it so every completed order publishes an event, a settlement consumer aggregates per seller, and a payout is created once the aggregate crosses the threshold. The hard part was idempotency: the consumer can see the same event twice, so I keyed every settlement line on the order id and made the insert conditional. We considered just running the batch more often, but that would not have fixed the failure mode, it would have made it four times a day instead of once. After the change the seller payout delay went from about twenty-six hours to under two.

Q3Can you explain how you handle duplicate or out-of-order settlement events to maintain accurate seller balances?

2srelevance · specificity · structureFeedback + model answer

What we heard

We had a double-settlement incident where a retry paid some sellers twice. I was on call. The immediate fix was to stop the payout job, then I reconciled the payouts against the ledger to find the affected sellers, which was about three hundred of them. Root cause was that a timeout on the bank API was being treated as a failure, so we retried, but the first call had actually succeeded. I fixed it by making the payout request idempotent with a reference id the bank honours, and I wrote the runbook the team still uses.

Q4Describe a time you introduced a new caching strategy to improve payment processing latency and how you measured its impact.

5srelevance · specificity · structureFeedback + model answer

What we heard

For the refund API, p95 was around 800 milliseconds. I profiled it and most of the time was a synchronous read against the ledger to check the refundable amount. I added a Redis cache for the refundable balance that gets updated when the ledger changes, and the read became a cache hit most of the time. That brought p95 down to about 140 milliseconds. The trade-off is that the cache can be briefly stale, so the refund service still validates against the ledger before committing, the cache only makes the happy path fast.

Q5What eviction policy did you choose for the Redis cache of refundable balances, and why?

2srelevance · specificity · structureFeedback + model answer

What we heard

Honestly, I have not built a reconciliation system from scratch. At Swiftcart there is a nightly reconciliation job that someone else owns, and I have used its output during incidents. I know the shape of the problem, matching our ledger against the bank statement and flagging mismatches, but I have not designed one. I would want to learn that here.

Try it again

You scored 74. The questions change on a second attempt and your weak spots do not, so the same gaps come back where you can see them coming.

03 / 04 Delivery

◷

How it sounded

Speech statistics and dimension averages are arithmetic over the answers above.

Speech Analysis

Counted from the transcript above, not from your audio.

Filler Words

2 (0.4%)

Breakdown

“actually” ×1 — “…e retried, but the first call had actually succeeded. I fixed it by making t…”

“right” ×1 — “…payments and orders at Swiftcart. Right now I own the settlement service…”

Some of these are hesitation and some are ordinary speech — the count cannot tell them apart, so the phrase is shown and you judge.

2 filler words in 457 words of transcript.

Dimension averages

Relevance (40%)75/100

Specificity (35%)77/100

Structure (25%)73/100

Top improvement areas

Answer the exact Redis question first, then explain whether the cache is rebuildable and how your policy prevents eviction from causing incorrect refundable-balance decisions.

04 / 04 Next moves

⚑

Before the next one

A working list for the next few days.

Personal Improvement Plan

Get a personalized 7-day plan based on your mock performance.

Do one mock interview. If the report does not give you one thing to change before your next interview, write to us within three days and tell us what it missed. We refund in full.

After your free mock

This is three questions. A real interview is twenty.

Your free mock gives you three questions, scored, with a model answer on your weakest. The pass gives you the rest.

Unlimited mocks, up to twenty questions each

A model answer on every answer, not just the weakest

Pre-game briefing for the company and the round

Your resume read by seven hiring systems

Tailoring, cover letters and the LinkedIn optimiser

Speech analysis: pace, filler words, where they fall

SAVE ₹1,497~~₹3,996~~₹2,499for a month. What four weekly passes would cost.Get the pass ₹999for a week. If your interview is this week.Get the pass

Do one mock interview. If the report does not give you one thing to change before your next interview, write to us within three days and tell us what it missed. We refund in full.

Your first mock is freeThree questions, scored, with a model answer on your weakest

No card needed

03How it works

Speak. Get scored. Walk in ready.

2 min01

Pick the role

Paste the posting, or choose from your applications. The interviewer reads it before it asks anything.

15 min02

Answer out loud

Three questions free, up to twenty on the pass. Voice, with follow-ups that respond to what you said. No camera.

Instant03

Read what was heard

Your words on the page, structure ticks on each answer, a score on relevance, specificity and structure.

Again04

Fix it and rerun

The questions change on a second attempt. Your weak spots do not, so the gap comes back where you can see it coming.

04What is different

Honest by construction.

Scores that hold still

The rubric is fixed, so the same answer scores the same on every run. A second attempt measures you, not the mood of a model.

Your answer, not a better one

"Say it this way" is built from what you actually said. Not a model answer from somebody else's career. Yours, said properly.

Stop whenever you like

Answer one question or twenty, you still get a report for what you did. Anything we could not measure says so.

05What it looks like in use

The gaps come back where you can see them coming.

Attempt 1 · trade-off missing on 3 of 566

Attempt 2 · trade-off missing on 1 of 574

Attempt 3 · every answer names the decision83

Company patterns loaded · Razorpay engineeringready

Weakest answer rewritten in your wordsQ4

06Only here · six things no other tool in India does

Built to be honest, which turns out to be rare.

Every one of these exists because a lie was found and removed. They are not features. They are rules.

Reads the layout of your file

Columns, tables, fonts, page count, contact placement, across seven hiring systems. Every other scanner reads the text and calls it done.

Never invents a fact

Dates, employers, metrics, skills. If it is not in your document it does not appear, and the removal is named so you can add it yourself.

Scores that hold still

The rubric is fixed. The same answer scores the same on every run, so a second attempt measures you, not the mood of a model.

Rewrites your answer, not a better one

“Say it this way” is built from what you actually said. Not a model answer from someone else’s career.

Stop halfway, still get a report

Answer two questions or twenty. Anything we could not measure says “not enough to measure” instead of printing a number.

Absence is always stated

Salary not listed. Page count not measured. Company pattern unknown. Never a green tick nothing earned.

07Pricing

You have an interview. Pick how long you need.

The same product either way. A week if it is Thursday, a month if the hunt has just started.

No card needed

Free

Do a real mock, check your resume, and read a full report before you decide.

₹0always

One mock interview, three questions, scored, with a model answer on your weakest
Unlimited ATS structure checks, all seven systems
One keyword match against a job posting
Resume builder, four templates, manual editing
Download with a ClearRound footer
Eight salary calculators and the negotiation scripts
A full sample of every paid report, start to finish

Start my free mock

Everything, for the time you need

The pass

The same product either way. The only choice is how long your hunt lasts.

Unlimited mock interviews, voice, up to twenty questions
Feedback and a model answer on every question, not just the weakest
Unlimited keyword matching against any posting
Resume tailoring, cover letters and STAR stories
Pre-game briefing for the company and the round
LinkedIn optimiser, prep reports, clean exports

SAVE ₹1,497~~₹3,996~~₹2,499 for a monthWhat four weekly passes would cost. Thirty days, paid once.₹999 for a weekSeven days, paid once. For an interview this week.

THE GUARANTEE

Do one mock interview. If the report does not give you one thing to change before your next interview, write to us within three days and tell us what it missed. We refund in full.

No form. One email within three days of the payment, saying what it missed.

08Fair questions

Things worth knowing before you start.

Is this just a chatbot asking questions?

No. The interviewer follows up on what you said, the scoring runs on a fixed rubric so the same answer scores the same twice, and the rewrite is built from your own words. A general chatbot cannot tell you that your fourth answer named a stack but not a decision.

What if my microphone does not work?

We check it before the interview starts and tell you plainly if we cannot hear you. If it still fails, type your answers instead. The report is the same.

What if I stop halfway through?

You still get a report for the questions you answered. Anything we could not measure from two answers says so rather than printing a number we cannot stand behind.

Is the free mock a trial?

No. It is one mock, three questions, once, and there is no card and no expiry on it. What the pass adds is unlimited mocks, twenty questions each, and a model answer on every answer rather than only your weakest.

Is ₹999 not expensive?

It is a week of unlimited scored practice. A career coach charges three times that for two hours and never hears you speak twice. If it does not help, you get it back.

The first time you say it out loud should not be in front of someone deciding whether to hire you.

Three questions this evening. Six minutes, scored, out loud, free.

Start my free mock Read the report again