Statistics questions in analyst interviews check whether you can use numbers sensibly, not whether you can derive formulas. Short, clear, example-based answers work best.
Descriptive statistics
Mean vs median: when would you use each?
The mean is sensitive to extreme values; the median is not. For skewed data such as incomes, order values or delivery times, the median better describes a typical case. Give an example: one very large order pulls the mean up but barely moves the median.
What is standard deviation?
A measure of how spread out values are around the mean. Small: values cluster closely. Large: values vary a lot.
What is variance?
The average squared distance from the mean; standard deviation is its square root, which brings it back to the original units.
What are percentiles and quartiles?
The pth percentile is the value below which p% of observations fall. Quartiles split data into four parts (25th, 50th = median, 75th). The interquartile range (Q3 โ Q1) measures the spread of the middle half.
How do you detect outliers?
Box plots, the 1.5 ร IQR rule, z-scores, or simply plotting the data. Then investigate: data error, rare genuine event, or the most important finding.
Distributions
What is a normal distribution?
A symmetric, bell-shaped distribution where mean, median and mode are equal. Roughly 68% of values lie within one standard deviation of the mean and about 95% within two.
What is skewness?
Asymmetry. Right-skewed (positive) data has a long tail of high values, so the mean exceeds the median. Many business metrics are right-skewed.
What is the central limit theorem?
The average of many independent samples tends towards a normal distribution, even when the underlying data is not normal. This is why many tests on averages work.
Relationships
Correlation vs causation
Correlation means two variables move together; causation means one changes the other. A third factor, coincidence or reverse causation can create correlation without causation. Controlled experiments are the usual way to establish causation.
What does a correlation of โ0.8 mean?
A strong negative linear relationship: as one rises, the other tends to fall.
Inference and testing
Population vs sample
The population is everyone of interest; a sample is the subset you measure. Good samples are representative and large enough.
What is a p-value?
The probability of seeing a result at least as extreme as yours if the null hypothesis (no real effect) were true. A small p-value suggests the result is unlikely to be chance alone. It is not the probability that the hypothesis is true, and it does not measure how large or important the effect is.
Type I vs Type II error
- Type I (false positive): concluding there is an effect when there is not.
- Type II (false negative): missing a real effect.
What is a confidence interval?
A range that, under repeated sampling, would contain the true value a stated percentage of the time (for example 95%). Wider intervals mean more uncertainty.
How does an A/B test work?
Randomly split users into two groups, show each a different version, measure a pre-chosen metric, and test whether the difference is larger than chance would explain. Decide the metric and sample size before starting, and do not stop early just because the result looks good. See our A/B testing guide.
Practical questions
- "Average order value went up but revenue went down. How?" (Fewer orders.)
- "A city has the highest complaint count. Is it the worst?" (Compare rates, not counts.)
- "Conversion improved in every segment but fell overall. How?" (Simpson's paradox: the mix of segments changed.)
Prepare
Read statistics for data analysts, compute summaries yourself in NumPy, and see how statistics drives model evaluation on the metrics page.
Practise what you just read
Reading helps; writing the code yourself is what sticks. Start here: