Data Analyst case study round: from question to recommendation

You are typically given a business scenario and either a dataset or summary numbers. A common format is a take-home followed by a presentation, or a live 45-minute discussion. Either way the interviewer is watching a sequence.

• Framing: can you restate the business question and name the decision it informs?
• Metric choice: do you pick metrics that map to the goal rather than whatever is easy to compute?
• Analysis: do you segment, compare and check for confounders?
• Recommendation: do you finish with a specific, caveated action?

The output that matters is the recommendation, but the score comes from the path you took to reach it.

Scenario: 'Checkout conversion dropped 8% last month. Investigate.' A strong candidate does not immediately blame the checkout page.

Strong approach: They decompose conversion into steps (view to cart, cart to checkout, checkout to paid), then segment by device, region, traffic source and new-vs-returning. They notice the drop is concentrated on mobile new users from one campaign, form a hypothesis, and quantify the impact. They close with: 'Most of the drop traces to a mobile payment error on one campaign; I recommend fixing the payment step and re-running an A/B check, expected to recover the bulk of the loss.'

Weak approach: They eyeball the total, guess it is 'seasonality', and offer no segmentation or hypothesis. The answer is plausible but untested, and the interviewer cannot see any analytical rigour.

Behind the scenario, the round is a proxy for how you would operate unsupervised. Most Data Analyst / Data Science Jobs at growth-stage companies expect this operate-unsupervised muscle from day one.

• Prioritisation: do you chase the biggest driver first instead of the most interesting one?
• Scepticism: do you sanity-check totals, question data quality and consider sample size?
• Communication: can a non-technical stakeholder follow your logic and act on it?

Weak candidates over-index on producing charts; strong candidates use the minimum analysis needed to make a defensible decision, and say what they would check next with more time.

Build a repeatable structure so you never freeze on a blank problem.

• Practise a simple loop: clarify the goal, state the metric, segment, hypothesise, quantify, recommend.
• Take public datasets and give yourself 40 minutes to answer a business question end to end.
• Rehearse the presentation aloud; the recommendation should be sayable in two sentences.

Since half the score is delivery, rehearse speaking the case, not just writing it. A free AI voice mock interview on InterviewPrep can generate a case-style prompt from your CV and a target job description, then score your structure, pace and filler words so your narrative sounds decisive rather than hesitant.