Tell me about a time you used data to solve a problem

• That you start from the question and not from the data. Weak candidates describe a file and tools. Good ones describe what they wanted to know and why.
• Your assumptions. Every real analysis rests on incomplete data. Saying what you assumed, and how you checked that it held, is what separates analysis from calculation.
• Your ability to simplify. If your explanation needs statistical jargon or the names of three software packages, you do not yet have the expected reflex.
• What was done with the conclusion. An analysis that changed no decision is an exercise.

1. 1

   State the problem in one sentence

   What decision had to be made, and by whom. That is what gives meaning to everything else.
2. 2

   Explain your starting point and assumptions

   What you expected to find, the data you had, what was missing and what you assumed in order to move forward.
3. 3

   Describe the approach in everyday language

   Two or three steps, with no tool or formula. Mention the moment a result surprised you and what you did to check it.
4. 4

   Give the conclusion and its effect

   What you recommended, what was decided, and the limit of your analysis you would flag today.

"During my internship at a fast food chain, the regional director wanted to know whether to stay open later in the evening. Everybody had an opinion, but nobody had looked at the numbers. My starting hypothesis was that the last hour did not cover its staff costs.

I had hourly till receipts over a year for twelve restaurants, but no costs by time slot. So I assumed the same minimum staffing everywhere to keep a restaurant open, and checked that assumption with three managers. The result surprised me: the last hour was profitable in restaurants near a station and clearly loss-making elsewhere. I first suspected an error, so I redid the calculation restaurant by restaurant..."

This example is made up: it shows the tone and the construction, not a script to memorise. Your answer has to come from your own experience or from the data the interviewer gives you.

• Describing the tool instead of the reasoning: the formulas, the programming language, the size of the file. The interviewer wants to know how you think.
• Presenting an analysis with no assumption and no doubt. It suggests you did not see the limits of your data.
• Forgetting the conclusion. Many candidates describe the method at length and stop before saying what they found and what was done with it.
• Choosing purely academic work, with no stakes and no audience, when a real situation, even a modest one, exists in your background.