Data & Analytics: interview questions and learning guide

SQL for analysis

Data sits in tables like rows in many notebooks. SQL is how you ask the notebooks questions.

SQL reads and summarises data stored in tables. SELECT picks columns, WHERE filters rows, JOIN links tables on a shared key, and GROUP BY summarises rows into totals and counts. HAVING filters those groups. Window functions rank or compare rows without collapsing them. Clean, readable queries and checking row counts after each join prevent wrong numbers.

Interview tip: Say how you check your answer: row counts, a small sample and a second method.

Dashboards and data storytelling

A chart is a sentence in pictures. If it needs a paragraph to explain, redraw it.

A dashboard shows a few key measures so people can decide quickly. Start with the question and the audience, then pick the simplest chart: lines for time, bars for comparing, tables for exact values. Use clear titles, consistent colours and a small number of KPIs. Always label units and the date range. Check numbers against the source before sharing, and note where the data comes from.

Interview tip: Show that you start from the decision the person must make, not from the chart type.

Statistics and insight basics

Averages can lie. Always ask how the numbers are spread and who is missing.

Statistics helps you describe data and decide how much to trust it. Mean and median describe the middle, and the median is safer when a few large values exist. Spread shows how varied the data is. Correlation shows two things move together but does not prove one causes the other. A fair test compares groups that differ in only one thing, and sample size matters. Report ranges and caveats, not just one number.

Interview tip: Never say a number proves something unless you explain the test and the sample.

WHERE filters rows before grouping. HAVING filters the groups after aggregation, for example totals above a limit.

First check the data and its source for errors, then split by time, region and product to find where the drop sits, and test likely causes with evidence before reporting.

Compare row counts and totals with the source, check a small sample by hand and cross check with a second method or an earlier report.

When a few extreme values pull the mean away from what is typical, such as salaries or order values, the median gives a more honest middle.

Group by the columns that should be unique and keep groups with COUNT greater than one, or use ROW_NUMBER over a partition.

To summarise data by categories quickly: choose rows, columns and a value with sum or count.

Mean is the average, median the middle value and mode the most frequent. Median resists outliers.

Find out why they are missing, then remove, fill with a sensible value or flag them, and record the choice.

A clear question, a few key numbers, simple charts, consistent colours, filters that help and a note on data freshness.

A primary key uniquely identifies a row. A foreign key points to a primary key in another table to link them.

Correlation measures how two variables move together. It does not prove one causes the other; other factors or chance may explain it.