Interview prep for product and data analysts

Paste a job posting.Get ready for its interviews.

A practice company with its own database, an AI interviewer that pushes back, and code that checks your numbers.

Free, no sign-in · Pro $29, paid once

←Prepare for a job

Paste the job. Get the prep.

Product Analyst — Parsnip (payments app, New York). Own checkout metrics: conversion, retention, LTV, CAC payback. Design and read A/B tests with PMs. SQL: cohorts, window functions. Python / Pandas. Explain findings to leadership. 231 / 12,000

Build my prep⌘↵

Your plan will include
✓Mock interviews, stage by stageProduct, A/B, SQL, Python · a score and 3 fixes
✓SQL, Pandas and Python problemsEasy to hard, with hints
✓A week as the analyst5 days at Carrot, a practice company · 2–3 h
✓“Am I ready?” reportEach requirement: ready, almost or gap

Your prep

Product Analyst · Parsnip

Top skills: Product metrics · A/B testing · SQL

Your plan 5 steps · in order

1Mock interview · Product case AI“Raise the referral bonus?” · ~18 minStart
2Mock interview · A/B testing AI“Carrot+: the first month free” · ~18 minafter step 1
3Mock interview · SQL AI“Do customers come back after a late delivery?” · ~18 minafter step 2
46 problems · SQL and Pandas codeA/B testing, statistics, cohorts · plus a quiz~40 min
5Work week, tuned for this job code5 days · 2–3 h · metrics, A/B testing, SQLafter step 4

0 of 6 readyOpen Readiness

←Mock interviewAISQLFind the bugsCategory revenue for August06:40 / 25

ProblemInterviewerCriteria

Our AI assistant wrote this query. The categories add up to about 14 times August’s revenue.

The JOIN repeats each order’s total on every line item, so revenue is inflated.

And <= '2026-08-31' ends at the start of Aug 31, so that day is lost.

Why not just SUM(DISTINCT o.amount_usd)? The repeats go away.

No — orders with equal totals would merge, and each category still gets the whole order.

PostgreSQLSubmit fixed query

1SELECT oi.category,

2 COUNT(DISTINCT o.order_id) AS orders,

3 SUM(o.amount_usdoi.quantity * oi.price_usd) AS revenue_usd

4FROM orders AS o

5JOIN order_items AS oi

6 ON oi.order_id = o.order_id

7WHERE o.status = 'delivered'

8 AND o.created_at >= '2026-08-01'

9 AND o.created_at <= '2026-08-31'< '2026-09-01'

10GROUP BY oi.category

The query ran: 8 rows

category	orders	revenue_usd
Beverages	1195	11257.98
Dairy	1794	19864.40
Household	791	15229.52
Meat & seafood	1713	68507.26
Pantry	1407	18519.88
Prepared foods	922	22100.92
Produce	2233	36126.14
Sweets & snacks	1016	14607.22

How it works

Four steps, one job posting

Every step below is built from the posting you paste.

1of 4

Paste the job. Get a plan for it.

Mock interviews, problems and a practice work week, picked from what the posting asks. No posting? You get a general plan.
What happens to my posting
- The requirements are picked out right in your browser.
- If AI reads the posting, emails, phone numbers and links are hidden first.
- We don’t keep the text.
Paste the job description here (optional)
Build my prep →Free, no sign-in
2of 4

Rehearse the interview.

An AI interviewer asks follow-ups and challenges your answers. You get a score out of 100 and your top 3 fixes.
More about the interview
- Product, A/B testing, SQL or Python. A short one takes 15–20 minutes, a full one about 40.
- In SQL and Python you write code in a shared editor.
- A practice verdict on your answers, like an interviewer’s, then a debrief with a cheat-sheet card for each gap.
Example · A/B testing

InterviewerConversion is up 3%. The PM wants to ship. Your call?

YouFirst I’d check that the groups really split 50/50.

InterviewerThe PM says the split doesn’t matter if the effect is big. Agree?

72/100· your top 3 fixes
3of 4

Work a week as the analyst.

Five days at Carrot, a practice company: emails with deadlines, SQL on its database and an A/B test to decide on.
More about the week
- 2–3 hours in all, saved as you go. The traps are the ones from real work: a vague ask, a skewed split, a peek before the test ends, a colleague’s report to review.
- A wrong number comes back later in the week as an email.
- Friday: present your findings to the CEO AI
Example · Monday 8:41 AM

How many active users do we have?

ARAshley Rivera · MarketingDue Tue 12:00 PM

Hi! Can you pull how many active users we have? I need one number for Tuesday’s deck — any definition is fine.

The trap: “active” isn’t defined. Ask first, or pick a definition and say so.
4of 4

See where you’re ready.

Each requirement from the posting is marked ready, almost or gap. Code does the scoring, not AI.
When a skill counts as ready
- Only when two independent results agree — a mock interview and the work week, for example.
- One bar for everyone: what a senior analyst would do. New to it? Start with the easy problems and hints.
- A self-check in a simulator, not a certificate.
Example · another posting

Am I ready? · Data Analyst · Turnip

2 of 4 requirements ready
- SQLReady
- A/B testingReady
- DashboardsAlmost
- Explaining results to leadershipGap

The mock interview screen, SQL section: the task “Do customers come back after a late delivery” on the left, a shared PostgreSQL editor with a query on the right, and the result table below it. — **Mock interview, SQL section.** The task, a shared editor on PostgreSQL and the result. The interviewer is in the next tab.

The work week screen: Monday 9:00 AM at Carrot, an inbox with two requests and their deadlines, and an open email from Marcus Reed, the product manager, asking whether to ship the new cart. — **Work week, Monday 8:55 AM.** A PM wants a ship decision for Thursday’s committee — and his P.S. slips in a slice with p = 0.048.

The same email on a phone: New cart test results — do we ship? — **Work week, Monday 8:55 AM.** A PM wants a ship decision for Thursday’s committee — and his P.S. slips in a slice with p = 0.048.

Honest verdicts

Checked by code, not guessed by AI

Scores, practice verdicts and readiness are computed by code. AI plays the people and reads your writing — it never does the math.

codeCode checks

Your SQL result against the reference — on PostgreSQL, in your browser
A/B statistics: the split, the test you chose, the ship call
Every number in the work week and every practice answer

AIAI plays the people

The interviewer, the stakeholders and the CEO
Grades written answers by a rubric you can see, quoting your words for every point; code checks the quotes and adds up the score
Reads the posting once, with emails, phone numbers and links masked

Works without AI too: everything checked by code stays the same.

One bar for everyone

Ready means proven twice

Every answer is scored against what a senior analyst would do, whatever the posting says. A requirement counts as ready only when two independent results agree — a mock interview and a work week, for example. Practice alone isn’t enough. Junior? Same bar — just start with the easy problems and hints.

Who it’s for

An interview is coming up

You have a product or data analyst interview in a week or three. Paste the posting and spend that time on the interview’s sections, the work it describes and an honest list of gaps.

Aiming for the next level

Rehearse what a senior analyst owns: a ship call under pressure from a PM who wants a yes, choosing the main metric for a product area, impact in dollars.

You want practice that feels like work

You can solve puzzles. Now practice SQL, Pandas and A/B on one company’s data, with deadlines and a stakeholder who disagrees.

Not the best fit if you want a certificate or a promise of an offer. And if you’ve never written SQL, start with the basics first: Offer’s easy problems assume you can read a JOIN.

Compared

Compared with the usual mix

Problem sets, mock interviews with people — and Offer.

	Problem sets	Mock interviews with people	Offer
Your target job	Problem setsGeneric question lists	Mock interviews with peopleDepends on the interviewer	OfferMocks, practice and a week picked for the posting, plus a readiness report
The interview	Problem setsQuestions, no conversation	Mock interviews with peopleA real conversation — when you can book one, usually paid	OfferAn interviewer that asks back and pushes back, any time, in your browser
The work itself	Problem setsIsolated puzzles	Mock interviews with peopleYou talk about work you’ve done	OfferFive days at one company: emails, deadlines, SQL, an A/B test to call
Unclear asks	Problem setsThe prompt tells you everything	Mock interviews with peopleIf the interviewer plays along	OfferVague on purpose; each question to a stakeholder costs 15 game minutes
What’s checked	Problem setsWhether your output matches	Mock interviews with peopleOne person’s judgment	OfferNumbers, queries and statistics by code; writing by AI against a visible rubric

Offer has a problem set too: 220+ problems from easy to hard — SQL, Pandas and Python algorithms, “Find the bugs” tasks included — all checked by code, with hints if you’re stuck.

Pricing

Free to start

Problems, work weeks and your prep plan are free, plus one AI mock interview. Pro adds more AI when your interview is close.

Free

No card needed

One AI mock interview AI once you sign in: 15–40 min, a score and your top 3 fixes
220+ problems in SQL, Pandas and Python, checked by code
Work weeks at Carrot: 5 days of emails, SQL and an A/B test, checked by code
A prep plan and a readiness report for your job posting
Cheat sheet, plus “Rate the candidate”: score sample answers like an interviewer

Start free

Pro pass

$29 for 30 days

or $59 for 90 days, about $20 a month

One-time payment. No auto-renewal. Tax, if any, is shown at checkout.

Everything in Free, plus:

AI mock interviews AI — about 40 in 30 days, at your own pace
Work weeks with AI: ask stakeholders and your Carrot mentor anything, get feedback on write-ups, defend your call to the CEO

Interview within a month? Take 30 days. Still searching? Take 90.

Get 30 days — $29 Get 90 days — $59

Full refund within 7 days of buying, for any reason.

FAQ

Is Offer free?

Yes, to start: all practice, work weeks, your plan and report, plus one AI mock interview once you sign in. Pro adds more AI for 30 or 90 days, paid once.

Why a pass, not a subscription?

Most people prepare for a few weeks or months. Buy 30 or 90 days — nothing renews. Need more time? Buy again; the days add up. Your progress always stays.

Can I get a refund?

Yes, a full refund within 7 days of buying, for any reason. Just email support@offerbeta.com. Your progress stays. Refund policy

Is there a limit on AI?

Pro has no daily limit while its allowance lasts: about 40 full mock interviews in 30 days plus one AI work week each week, about 100 in 90 days. When it runs out, a small reserve gives you 1 mock a day. Anything started can be finished.

Do I need an account?

Not for anything checked by code — it runs in your browser, and your progress stays there. Sign in (it’s free) for your AI mock interview and to keep your progress across devices. You can also download it as a file.

Do I need AI?

No. Without AI, stakeholders reply from a script, the posting is read by keywords on your device, and every check by code works as usual. AI adds the interviewer, live stakeholder chat, the CEO defense and debriefs of your written answers.

Do I need a job posting?

No. Leave the field empty and you get the general product analyst program — the same kind of plan, built on the core requirements of the role.

What happens to the posting I paste?

We don’t keep the posting text — Offer keeps only the list of requirements it found. First the posting is read by keywords on your device. If AI is on, the text is also sent once to AI to pick out the requirements, with emails, phone numbers and links masked. Remove recruiter names before pasting: the mask doesn’t catch names.

What changes when I paste a posting?

The sections and cases of your mock interviews, the practice set, the cheat-sheet cards, and up to three of the week’s emails with its A/B scenario. The company in the week stays Carrot and the bar stays the same, so a score means the same thing for everyone.

What level does Offer assess?

One bar for everyone: Senior. Every interview answer and every work request is scored against what a senior product analyst would do. Your game level still grows step by step: a work week ends with a score from 0 to 100 — Junior, Mid-level at 65+, Senior at 85+. A posting’s level only changes which cases you get, not the bar.

Is the readiness report a certificate?

No. It’s a self-check in a simulator: it shows which requirements your results in Offer back up and where the gaps are. No tool can promise a job offer, and we don’t.

Which languages are available?

English and Russian, both complete. The English version is set in the US — Carrot, dollars, American cities; the Russian one has its own company, rubles and Russian cities.

Is Carrot a real company?

No. Carrot, its people and its data are fictional. The data is generated so that every question has a known right answer — that’s how code can check you. SQL runs on PostgreSQL right in your browser; Pandas and Python do too.

Got an interview coming up?

Paste the posting — or skip it — and take the first step today.

Build my prep Free to start · 1 AI mock interview included · Runs in your browser