Take home assignments are now a normal part of hiring for product, data and growth roles. They're also where good candidates lose offers by over delivering in the wrong places, missing the actual question, or running out of time. This guide covers what take homes look like for each role, how graders score them, and a structure that works, with full example outlines.
Why companies use take homes
A take home shows how you work on a realistic problem with time to think, which a 45 minute interview can't. It also lets the team compare candidates on the same task. From the company's side, a good take home answers three questions:
- Did you understand the real problem?
- Is your reasoning sound and your work correct?
- Can you communicate it clearly to the people who'd act on it?
That third point decides more take homes than candidates expect.
Common formats by role
| Role | Typical task | Usual length |
|---|---|---|
| Product manager | Write a short PRD or strategy memo for a feature, or diagnose a product problem and propose a plan | 2 to 4 hours, 2 to 4 pages |
| Data analyst | Analyze a CSV or database, answer business questions, build a chart or dashboard | 3 to 5 hours, plus SQL or notebook |
| Data scientist | Design or analyze an experiment, build a simple model, explain tradeoffs | 4 to 8 hours |
| Growth marketer | Plan a channel test or campaign with a budget, forecast results | 2 to 4 hours |
| Strategy and ops | Build a model in a spreadsheet and recommend a decision | 3 to 5 hours |
How graders actually score them
Most teams use a rubric with four or five areas. A typical one:
- Problem framing (25%): did you restate the goal, state assumptions and focus on what matters?
- Correctness and rigor (30%): are the numbers right? Did you check data quality? Are conclusions supported?
- Judgment (20%): did you make a clear recommendation and weigh tradeoffs?
- Communication (25%): could a busy executive get the answer in one minute?
Notice that nothing rewards length. A tight 2 page answer usually beats a 15 page deck.
A structure that works for any take home
- Answer first. Start with a three line summary: the recommendation, the key evidence, and the biggest risk.
- Restate the problem and your assumptions in a few bullets.
- Show the analysis in the order a reader needs it, with only the charts that change the decision.
- Recommend clearly, with what you'd do first and how you'd measure it.
- Risks and what you'd do with more time.
- Appendix for code, extra charts and data checks.
Example outline: PM take home
Prompt: "Our meditation app's day 30 retention fell from 22% to 18% over the last quarter. Diagnose the likely causes and propose a plan for the next two quarters."
Summary: "Retention fell mainly because new users from a paid social campaign churn twice as fast as organic users, and the new onboarding added a paywall before the first session. I'd move the paywall after the first completed session and tighten campaign targeting. Expected result: day 30 retention back to 21% to 22% within a quarter."
Sections: what changed in the quarter (releases, campaigns, pricing); retention by channel and cohort; the onboarding funnel before and after; three options compared on impact, effort and risk; the plan for two quarters with milestones; success metrics and guardrails; open questions.
Why it works: it separates mix effects (more low intent users) from product effects (the paywall), and it commits to a plan with numbers.
Example outline: data analyst take home
Prompt: "Here are 12 months of orders and customers. What's driving the slowdown in revenue growth, and what should we do?"
Approach:
- Check data quality first: duplicates, refunds, missing dates, test accounts. Note what you cleaned.
- Split revenue into customers times orders per customer times average order value. Find which part slowed.
- Look at cohorts: are new customers as valuable as older ones? Is repeat purchase falling?
- Segment by country, channel and product category.
- Recommend one or two actions with an estimated impact.
Deliverables: a one page summary, three charts, and a clean SQL or notebook appendix. Label every chart with what the reader should take from it, not just what it shows.
Example outline: growth take home
Prompt: "You have $50,000 to test a new acquisition channel for our B2B tool. Plan the test."
Answer shape: pick one or two channels with reasons; define the funnel from click to paying customer; set targets (cost per signup, signup to trial, trial to paid) from current benchmarks; compute the break even CAC from margin and expected lifetime; design the test with a budget split and stop rules; say what result would make you scale, change or stop.
The detail that separates strong answers: a clear stop rule decided in advance, so you don't keep spending on a channel that isn't working.
Mistakes that cost offers
- Ignoring the time limit. Spending 20 hours on a 4 hour task can read as poor prioritization, and it isn't sustainable.
- Burying the answer. If the recommendation is on page 9, many graders won't find it.
- No data checks. Not noticing duplicates or refunds in the data is an instant red flag for analyst roles.
- Too many charts. Every chart should support the decision.
- Using AI without understanding. Many teams now ask follow up questions about your take home, and some can see how you produced it. Use tools to speed up, but make sure every number and claim is yours.
Questions to ask before you start
- How much time do you expect this to take?
- Who's the audience: the hiring manager, the team, an executive?
- What format do you prefer?
- Can I make assumptions where the data is missing?
Asking these shows judgment and often gets you useful hints.
After you submit
Expect a follow up discussion where you walk through your work. Prepare a two minute summary, know which assumption you're least sure about, and be ready to say what you'd change with more time. That conversation often matters as much as the document.
How to split your time
| Share of time | What to do |
|---|---|
| 10% | Read the prompt twice, write down the real question and your assumptions |
| 15% | Explore and check the data, or research the product |
| 40% | Do the core analysis or thinking |
| 25% | Write the summary, recommendation and the few charts that matter |
| 10% | Edit: cut, check numbers, read it as the busiest reader would |
Most people spend 80% on analysis and rush the writing. Graders feel the difference immediately.
What a strong opening summary looks like
Recommendation: move the paywall to after the first completed session and cap paid social spend on the two lowest quality ad sets. Why: day 30 retention fell from 22% to 18% because new paid social users churn twice as fast as organic users, and the paywall now appears before users experience value. Expected impact: retention back to 21% to 22% within one quarter. Biggest risk: short term revenue dips while trial behavior adjusts.
Four sentences, and the reader knows what to do, why, what to expect and what could go wrong.
Data checks worth stating
- Row counts before and after cleaning, and what you removed.
- Duplicates, test accounts and refunds.
- Date ranges, missing days and time zone assumptions.
- Outliers you kept or removed, and why.
Listing these in an appendix takes ten minutes and signals rigor more than any chart.
When a take home is a red flag
Most take homes are fair. Be cautious if the task is unpaid and clearly real production work, such as building a full feature they'll ship, if it has no time guidance and is very large, or if the company won't discuss it with you afterward. It's reasonable to ask for a time estimate or a smaller scope.
Questions
How long should a take home assignment take?
Usually 2 to 5 hours. Ask the company how long they expect, and stick close to it.
How are take home assignments graded?
Most teams score problem framing, correctness, judgment and communication. Clear recommendations up front matter more than length.
Should I use AI on a take home assignment?
Check the rules. Many teams allow tools but ask follow up questions, so every number and claim should be one you can explain.
What format should a data analyst take home use?
A one page summary with the answer first, a few labeled charts that support the decision, and a clean SQL or notebook appendix.
What's the most common take home mistake?
Burying the answer. Put your recommendation, key evidence and biggest risk in the first three lines.