Interview Questions
Answering like a professional.
Data interviews usually include SQL, statistics, and a business case. The case is where people struggle. Here's how to answer calmly and clearly.
- SQL, statistics, a business case
- The case is the hard part
- Answer calmly
Revenue dropped ten percent. Why?
Don't guess. Follow a method. First, check that the data is right. Is it a real drop, or a tracking bug? Then split it, by region, product, and customer type, to find where the drop is. Next, split revenue into orders and average value. Then look for causes, such as prices, competitors, or the season. Finally, propose a way to test your idea.
How would you test a new feature?
Give a full answer. State the hypothesis, and one main metric. Calculate the sample size. Randomise users. Run for full weeks. Don't peek. Check that groups are balanced. Then read the effect size and interval, and make a decision. If you can't randomise, mention difference in differences.
Explain a p-value to a manager
Keep it simple. If this change did nothing at all, a result this large would happen by chance only about two percent of the time, so we're fairly confident that it's real. Then add, here's the size of the effect, and here's the range. Managers care about the size, not the p value.
Summary
Follow a method, and keep it plain. That's how to shine. Next, your cheat sheet.
Key takeaways- Check, slice, explain, test
- Effect size before p-value
- Plain language wins