Data analyst take-home assignment: how to ace it

To ace a data analyst take home assignment, clarify the business question first, time-box your work, document every cleaning decision, and lead your submission with a one-page summary of three to five findings and recommendations. Keep code and charts clean and reproducible, state your assumptions, and practise walking the panel through it in ten minutes.

You cleared the screening call and the SQL round. Now the email arrives: "Please find attached a dataset. Share your analysis within 72 hours." For many data analyst roles, this take-home assignment is the round that decides the offer.

It is also the round where good analysts often lose out, not because their analysis is wrong, but because they spend 20 hours on charts nobody asked for and bury the answer on slide 14. This guide gives you a repeatable plan and a full worked example. For the question bank side of preparation, see our data analyst interview questions.

What are companies really testing in a take-home assignment?

The dataset is usually simple on purpose. The panel is not checking whether you know pandas or DAX syntax. They already tested that. They are checking how you think when the problem is messy and nobody is there to guide you.

• Business sense: do you understand what decision the analysis supports?

• Data judgement: do you notice duplicates, odd values and missing data, and handle them sensibly?

• Prioritisation: can you pick the three findings that matter out of the twenty you could report?

• Communication: can a busy manager understand your conclusion in under two minutes?

• Reliability: is your work clean, reproducible and delivered on time?

How should you start before you touch the data?

Read the brief twice and write down, in one sentence, the business question. If the brief is vague, it is completely acceptable, and usually impressive, to send one short clarifying email.

Thanks for sharing the assignment. Before I begin, two quick questions: should I focus on the last 12 months or the full date range, and is the audience for the summary the product team or leadership? I will assume the full range and a leadership audience if I do not hear back by tomorrow morning.

Notice the last line. It shows initiative without blocking yourself. Then set a time budget. If they say "this should take around four hours", respect it. Spending three days signals poor prioritisation, and many companies ask how long you spent.

What is a good step by step plan for the assignment?

Here is a plan that works for a typical four to six hour assignment. Adjust the times, but keep the order.

1. Scope (20 minutes): write the business question, the audience and three hypotheses you want to test.

2. Profile the data (30 minutes): row counts, date ranges, nulls, duplicates, unique values in key columns and obvious outliers.

3. Clean and log (45 minutes): fix issues and write each decision in a short cleaning log, for example "Removed 212 duplicate order IDs; kept the latest timestamp."

4. Analyse (90 minutes): test your hypotheses first, then explore. Stop exploring when you have enough to answer the question.

5. Build visuals (45 minutes): four to six charts, each with a title that states the finding, not just the metric.

6. Write the summary (40 minutes): one page with the answer, three to five findings, recommendations and limitations.

7. Review (20 minutes): rerun everything from scratch, check numbers match across the summary and charts, and fix typos.

What does a strong submission look like? A worked example

Meera, a data analyst with three years of experience at a Wipro analytics practice, applied to an e-commerce startup. Her brief: "Here are 18 months of orders and returns. Our return rate is rising. What is going on, and what should we do?"

Step 1: Scoping and hypotheses

She wrote the question as: "Why has the return rate risen over the last six months, and which actions would reduce it the most?" Her hypotheses were that returns were concentrated in a few categories, that a new courier partner was causing damage, and that a recent sale had attracted low-intent buyers.

Step 2: Profiling and cleaning

Profiling showed duplicate order rows from a system migration, return dates earlier than order dates in a small number of rows, and missing city values for some orders. She removed the duplicates, excluded the impossible dates and kept missing-city orders in totals but out of the city breakdown. Each decision went into a six-line cleaning log at the end of her notebook.

Step 3: Findings

Notice that two of her hypotheses turned out to be wrong, and she said so clearly. Panels value that more than an analysis that conveniently confirms everything.

Step 4: The one-page summary

Summary: The rise in returns is driven mainly by footwear, and specifically by size-related returns after a new brand was added in March. Courier changes and the sale period are not major factors. I recommend auditing that brand's size chart and adding a fit guide on footwear pages, then tracking footwear returns weekly for two months. Limitations: I excluded a small number of rows with invalid dates, and I could not check product page changes because that data was not included.

Which tools and format should you use?

Use what the brief asks for. If it does not say, choose the tools the team uses, which you can often spot in the job description. Common, safe formats are:

• A short slide deck or PDF (six to ten slides) for the summary and charts.

• A clean SQL file or Jupyter notebook with headings that match your slides, so reviewers can follow the logic. Our SQL developer questions are useful revision for writing tidy queries.

• A Power BI or Tableau file if the role is dashboard-heavy. Keep it to one or two pages, not ten. The Power BI developer questions cover the modelling choices interviewers ask about.

• A README with how to run the code, the time you spent and your assumptions.

On AI tools: some companies allow them for take-homes and some do not. Read the instructions, ask if unclear, and be ready to explain every line of your work in the follow-up call. The analysis has to be yours.

How do you present the assignment in the follow-up call?

Most take-homes end with a 30 to 45 minute review call. Treat it as a presentation to stakeholders, not a code walkthrough. Aim for ten minutes of presenting and the rest for questions.

1. Open with the business question and your answer in one sentence.

2. Walk through the three or four key findings, one chart each.

3. Share your recommendations and what you would measure next.

4. Briefly cover data issues and how you handled them.

5. Invite questions and be ready to open your code if they ask.

Expect challenge questions such as "Why did you exclude those rows?", "What if the new brand was also discounted?" or "What would you do with one more week?" Prepare short answers. Practising this out loud in a mock interview makes a big difference, because explaining analysis verbally is a different skill from doing it.

What mistakes cost candidates the offer?

• No clear answer. Twenty charts and no conclusion. Always lead with the answer.

• Silent data cleaning. Changing data without telling the reader makes your numbers untrustworthy.

• Chart titles that just name metrics. "Returns by category" is weaker than "Footwear drives most of the rise in returns."

• Overbuilding. A predictive model nobody asked for, while the basic question stays unanswered.

• Late or untidy submission. Broken file links and code that does not run undo good analysis.

Once you have a polished take-home you are proud of, it becomes a portfolio piece (with any confidential company data removed). Keep applying in parallel with Auto-Apply for data analyst jobs, where every application waits for your approval. When you are ready to practise the review call or the next interview round, try InterviewPlan AI free with 5 live minutes and 100 AI credits, no card needed.

Frequently asked questions

How long should I spend on a data analyst take home assignment?

Stick close to the time the company suggests, usually four to eight hours. If none is given, aim for one focused day. Spending far longer can signal weak prioritisation, and some companies ask how long you spent. Note your actual time in the README.

What should a take home assignment submission include?

Include a one-page summary with your answer, key findings and recommendations, four to six charts with finding-style titles, clean and reproducible code or a dashboard file, a short cleaning log and a README listing your assumptions, limitations and how to run your work.

Is it okay to ask questions about the take home assignment?

Yes. One short clarifying email is professional and often impressive. Ask about scope, audience or ambiguous fields, and state the assumption you will use if they do not reply in time so you are not blocked waiting for an answer.

Can I use AI tools for a take home assignment?

Only if the company allows it. Read the instructions and ask if the rules are unclear. Whatever tools you use, the analysis and conclusions must be your own, and you should be able to explain every step and every line of code in the follow-up call.

How do I present my take home assignment in the interview?

Treat it like a stakeholder presentation. Start with the business question and your answer, walk through three or four findings with one chart each, share recommendations, mention data issues briefly and leave most of the time for questions. Keep it to about ten minutes.