Data analyst statistics interview questions

As asked

A stakeholder always asks for the average. When is the average the wrong summary, and what would you report instead?

Sample answer outline

The mean is distorted by skew and outliers, so for salaries, house prices, or order values the median is usually more honest about the typical case. Explain skewed distributions plainly, mention that the mode matters for categorical or bimodal data, and that reporting a spread such as the interquartile range stops a single number from hiding the shape. Tie the choice back to the decision the stakeholder is making.

Expect these follow-ups

• How would you explain skew to someone with no stats background?
• What does the gap between mean and median tell you?
• When is the mean genuinely the better choice?

As asked

Ice cream sales and drowning rates both rise in your data. A stakeholder wants to act on the correlation. How do you respond?

Sample answer outline

Explain the confounder, here hot weather driving both, and why correlation alone cannot justify an intervention. Offer ways to get closer to causation: control for the confounder, look for a natural experiment, or run a controlled test. Keep the tone collaborative rather than lecturing, and propose a concrete next step that would actually answer the stakeholder's underlying question.

Expect these follow-ups

• How would you design a test to check for real causation here?
• What other confounders might hide in marketing spend data?
• How do you push back without sounding obstructive?

As asked

You report that conversion lifted by 3 percent, plus or minus 2 percent. Explain to a marketing manager what that range means and why it matters.

Sample answer outline

Describe the interval as the plausible range for the true effect given the sample, not a guarantee, and explain that a range crossing zero means you cannot be confident there was any lift at all. Connect width to sample size so the manager understands why more data narrows the range. Avoid the technical misreading and focus on what action the range supports or blocks.

Expect these follow-ups

• What happens to the interval if we run the test for another week?
• Why is a point estimate alone dangerous to act on?
• When is a wide interval still good enough to decide?