Data Analyst Interview Questions Asked by Delhi NCR Employers

Typically solved using a correlated subquery, a window function (DENSE_RANK), or by ordering distinct salaries and skipping the first.

INNER JOIN returns only matching rows from both tables; LEFT JOIN returns all rows from the left table plus matches; FULL OUTER JOIN returns all rows from both tables, with NULLs where there is no match.

WHERE filters rows before grouping; HAVING filters aggregated groups after GROUP BY.

Use PARTITION BY category ORDER BY sales DESC, then filter where the ranking column ≤ 3.

A CTE is a named temporary result set defined with WITH. It improves readability and reuse, especially when the same logic is referenced multiple times or when nesting would become hard to read.

Use GROUP BY … HAVING COUNT(*) > 1 to find duplicates, then ROW_NUMBER() over the duplicate key ordered by an ID to keep the first row and delete the rest.

XLOOKUP can look up values in any direction, has a built-in if-not-found argument, defaults to exact match, and works across arrays — VLOOKUP is limited to rightward lookups and column index numbers.

Place Region in Rows, Month in Columns, Sales in Values, and group dates by month if needed.

SUMIFS/COUNTIFS accept a range plus criteria pairs, so you can sum or count rows that meet several conditions simultaneously (e.g., region = North AND month = Jan).

Power Query is used for repeatable data import and transformation. Unlike formulas, its steps are recorded and can be refreshed, making it far more scalable for recurring cleaning tasks.

Use a formula-based rule such as values beyond mean ± 2 standard deviations, or an IQR-based rule, and apply a colour highlight.

A calculated column is computed row by row and stored in the model; a measure is evaluated at query time based on filter context, so it is more dynamic and memory-efficient.

Define roles in Power BI Desktop using DAX filters on a user or region table, then assign users to roles in the Power BI Service.

Filters restrict data shown; parameters are user-controlled values that can drive calculations or filters; sets are custom groupings of members that can be used in calculations and comparisons.

Reduce the data model, use aggregations, remove unnecessary columns and visuals, use Import mode where possible, optimise DAX, and avoid high-cardinality relationships.

Leadership dashboards focus on trends, KPIs and exceptions at a summary level; operations dashboards focus on daily, actionable, drill-down metrics tied to specific tasks.

It's the probability of seeing a result this extreme if there were actually no real effect — a small p-value suggests the result is unlikely to be down to chance alone.

It states that the sampling distribution of the mean approaches normality as sample size grows, which is what allows us to use standard statistical tests even when the underlying data isn't normal.

Segment revenue by product, region, channel and customer cohort; compare against the same quarter last year and the previous quarter; check for pricing, mix, churn or volume effects; validate data quality; then isolate the largest contributors to the decline.

Randomly split users, define a primary metric and minimum detectable effect, run long enough to reach adequate power, and use a hypothesis test (usually a two-sample test) with a pre-set significance threshold such as 0.05.

Correlation means two variables move together; causation means one actually drives the other. A classic trap is assuming a marketing channel caused conversions when it merely captured users who would have converted anyway.

Use the STAR format and quantify the outcome wherever possible.

Separate the data from the interpretation — walk through assumptions, validate the source, and present alternatives rather than defending a single conclusion.

Cover how you profiled the data, documented the gaps, chose a handling method (deletion, imputation or flagging), and communicated the resulting limitations.

Connect a specific project to the sector's data problems, then show what you learned from it.

Questions on JOIN types (inner, left, full outer) and window functions (RANK, ROW_NUMBER) are among the most frequently asked, since almost every analytics role starts with pulling and shaping data from a database.

Yes, especially consulting firms (concentrated in Gurugram) and BFSI companies, where candidates are expected to structure an open-ended business problem, not just answer a technical question.

Both are common; the specific tool tested usually matches what the hiring company already uses internally, which is why Uncodemy's syllabus teaches both.

Not deeply, but enough to explain a p-value, describe a hypothesis test, and reason about correlation vs. causation in plain business language.