Data Analyst Interview Questions and Answers

A data analyst interview usually tests more than your ability to define technical terms. Interviewers want to know whether you can clean messy data, write SQL queries, work with spreadsheets and visualisation tools, interpret statistics, and explain your findings in a way that supports a business decision.

For freshers, questions often focus on data analysis fundamentals, Excel, SQL, data cleaning and basic statistics. Candidates with experience may also be asked about Python, dashboards, stakeholder requirements, business metrics and situations where the data is incomplete or misleading.

This guide covers data analyst interview questions and answers for freshers and experienced candidates, including SQL, Excel, statistics, Python, Power BI, data cleaning and scenario-based questions.

Basic Data Analyst Interview Questions and Answers

1. What does a data analyst do?

A data analyst collects, cleans, analyses and interprets data to answer business questions and support decisions. The role can include querying databases, validating data quality, identifying trends, creating reports and dashboards, and presenting insights to stakeholders.

A typical data analyst may use SQL to retrieve data, Excel for quick analysis, Python or R for larger or more complex datasets, and tools such as Power BI or Tableau for reporting and visualisation.

2. What skills are required to become a data analyst?

A strong data analyst usually needs a combination of technical, analytical and communication skills.

• SQL: querying, filtering, joining and aggregating data stored in relational databases.
• Excel: formulas, PivotTables, lookups, conditional aggregation and data cleaning.
• Statistics: mean, median, distributions, correlation, hypothesis testing and basic probability.
• Data visualisation: creating understandable reports and dashboards in Power BI, Tableau or similar tools.
• Python or R: useful for data manipulation, automation and more advanced analysis.
• Business understanding: connecting analysis with the question or decision the organisation is trying to address.
• Communication: explaining findings clearly to both technical and non-technical stakeholders.

3. Which tools are commonly used by data analysts?

The tools depend on the organisation and type of analysis, but commonly used options include:

Read Also: Power BI Interview Questions and Answers 2026

4. What are the main steps in the data analysis process?

A practical data analysis workflow usually includes the following stages:

1. Define the problem: understand the business question and the decision the analysis needs to support.
2. Collect the data: identify suitable databases, spreadsheets, APIs or other sources.
3. Clean and validate: check missing values, duplicates, incorrect data types, inconsistent categories and unusual records.
4. Explore the data: examine distributions, trends, relationships and potential anomalies.
5. Analyse: use SQL, statistics, spreadsheets or programming tools to answer the question.
6. Visualise and interpret: convert results into meaningful charts, tables or dashboards.
7. Communicate recommendations: explain what the findings mean and what action could be taken.

5. What is data cleaning?

Data cleaning is the process of identifying and correcting problems that can make a dataset unreliable or difficult to analyse. It does not simply mean deleting incorrect records.

Depending on the dataset, cleaning can involve:

• removing or investigating duplicate records;
• standardising categories and formats;
• correcting incorrect data types;
• handling missing values;
• checking outliers and impossible values;
• validating records against business rules; and
• checking whether joins or transformations changed row counts unexpectedly.

The objective is to produce data that is consistent and reliable enough for the intended analysis.

6. How would you handle missing values in a dataset?

I would first determine why the values are missing and how much of the dataset is affected. The correct treatment depends on the business context and the type of variable.

• Investigate the source: confirm whether the value is genuinely unavailable or missing because of a data collection problem.
• Measure the impact: check the percentage and pattern of missing data.
• Remove records: this may be reasonable when only a very small number of non-critical rows are affected.
• Impute values: depending on the situation, numerical values may be replaced using measures such as the median or model-based methods.
• Keep missing as a category: for some categorical fields, the fact that a value is missing may itself carry useful information.

A strong interview answer should also mention that you would document the method used because the treatment of missing data can affect the final result.

7. What is Exploratory Data Analysis (EDA)?

Exploratory Data Analysis is the process of examining a dataset before performing formal analysis or modelling. Its purpose is to understand the structure of the data and identify patterns or problems that may influence the analysis.

During EDA, an analyst may examine:

• summary statistics such as mean, median and percentiles;
• distributions of numerical variables;
• missing values and duplicates;
• outliers;
• correlations and relationships between variables; and
• unexpected patterns or data-quality issues.

8. How would you validate a dataset before analysing it?

I would validate the dataset at several levels instead of relying on a single check.

• Completeness: check whether required fields and expected records are present.
• Uniqueness: identify unexpected duplicate records or duplicate keys.
• Validity: confirm that dates, categories and numerical values fall within acceptable ranges.
• Consistency: make sure the same information follows consistent formats across datasets.
• Referential integrity: verify that related keys match correctly when datasets or tables are joined.
• Reconciliation: compare important totals or row counts with the source system where possible.

9. What is the difference between data analysis and data mining?

Data analysis is the broader process of examining data to answer questions, test assumptions, identify trends and support decisions.

Data mining focuses more specifically on discovering patterns, relationships or structures in large datasets, often using statistical or machine learning techniques.

10. How would you estimate the number of buses operating daily in Bengaluru?

This is a guesstimate or analytical thinking question. The interviewer is usually more interested in your assumptions and structure than in the exact number.

A strong answer could begin by clarifying whether the question refers to public buses only or all buses operating in the city. If public buses are the focus, I would estimate the answer using factors such as:

• population and daily public transport demand;
• percentage of commuters using buses;
• average passengers carried by one bus per trip;
• average number of trips completed by each bus per day; and
• fleet utilisation, since not every registered bus will operate every day.

I would then calculate an approximate fleet requirement and compare the estimate against another benchmark, such as the number of depots or the city’s registered public bus fleet. Clearly stating assumptions is more important than guessing a precise figure.

Advanced Data Analyst Interview Questions and Answers

11. What are some common challenges you may face during data analysis?

Common challenges include poor data quality, unclear business requirements, inconsistent definitions across teams, missing values, duplicate records, changing stakeholder expectations and combining data from multiple systems.

• Incomplete or inaccurate data: missing values, incorrect formats or inconsistent categories can affect the result.
• Multiple data sources: tables from different systems may use different identifiers, units or definitions.
• Unclear metrics: teams may interpret terms such as active user, revenue or conversion differently.
• Changing requirements: stakeholders may revise the original question after seeing early findings.
• Time constraints: analysts sometimes have to balance speed with accuracy and depth of analysis.

In an interview, it is better to explain one real challenge, how you diagnosed it, what action you took and how it affected the final analysis.

12. How should you explain a data analytics project in an interview?

A good project answer should focus on the problem, your approach, tools used and measurable outcome, rather than listing every task you completed.

1. Business problem: explain what question the project was trying to answer.
2. Data: mention the main data sources and important quality issues.
3. Approach: explain how you cleaned, explored and analysed the data.
4. Tools: mention relevant tools such as SQL, Excel, Python, Power BI or Tableau.
5. Insight: explain the most important finding.
6. Impact: state what decision, improvement or recommendation resulted from your analysis.

For example: “I analysed customer churn using SQL and Python, identified that customers with repeated support issues had substantially higher churn, and presented the finding through a Power BI dashboard. The analysis helped the team prioritise a retention campaign for high-risk customers.”

13. What is KNN imputation?

K-Nearest Neighbours (KNN) imputation is a technique used to estimate missing values based on observations that are most similar to the record containing the missing value.

The algorithm identifies the K nearest observations using a distance measure and estimates the missing value from those neighbours. For numerical variables, this may involve an average or weighted average; for categorical variables, the most common neighbouring category may be used.

KNN can preserve relationships in the data better than simply replacing every missing value with the mean, but it can become computationally expensive on large datasets and is sensitive to feature scaling.

14. What is VLOOKUP in Excel?

VLOOKUP searches for a value in the first column of a selected range and returns a corresponding value from another column in the same row.

The basic syntax is:

=VLOOKUP(lookup_value, table_array, col_index_num, [range_lookup])

For an exact match, the final argument is usually set to FALSE.

However, in newer versions of Excel, XLOOKUP is often more flexible because it can look both left and right, uses separate lookup and return arrays, and provides clearer handling when a match is not found.

15. What is a PivotTable in Excel?

A PivotTable is an Excel tool used to summarise and analyse large datasets without writing complex formulas. It allows users to group, filter and aggregate information quickly.

• Rows: categories displayed vertically.
• Columns: categories displayed horizontally.
• Values: measures such as sum, count, average or percentage.
• Filters: fields used to restrict the data included in the analysis.

For example, a sales analyst could use a PivotTable to compare total revenue by region, product category and month.

16. What is a waterfall chart and when would you use it?

A waterfall chart shows how a starting value changes through a sequence of positive and negative contributions until it reaches a final value.

It is useful for explaining:

• how revenue becomes profit after different costs;
• month-on-month changes in a business metric;
• budget increases and decreases;
• how individual factors contribute to an overall variance.

Choose a waterfall chart when the goal is to show how different components collectively explain a change, rather than simply comparing independent categories.

17. What is time series analysis?

Time series analysis studies observations collected over time to identify patterns and, in some cases, forecast future values.

A time series may contain:

• Trend: a long-term upward or downward movement.
• Seasonality: patterns that repeat at regular intervals.
• Cyclical movement: longer-term fluctuations that may not follow a fixed period.
• Noise: random variation not explained by the other components.

Data analysts may use moving averages, exponential smoothing or statistical forecasting models depending on the problem and data available.

18. What is machine learning, and how is it different from data analysis?

Machine learning is a branch of artificial intelligence in which algorithms learn patterns from data and use those patterns to make predictions, classifications or other automated decisions.

Data analysis is broader and often focuses on understanding what happened, why it happened and what the data means for the business. Machine learning is more commonly used when the objective is to predict or automate outcomes at scale.

For example, a data analyst may investigate why customer churn increased, while a machine learning model may estimate which customers are most likely to churn next month.

19. What is a data warehouse?

A data warehouse is a central system designed to store integrated, historical data from multiple sources for reporting, business intelligence and analytics.

Unlike operational databases that are mainly optimised for day-to-day transactions, a data warehouse is structured for analytical queries across large datasets.

Common characteristics include:

• data consolidated from multiple business systems;
• historical data retained over time;
• consistent formats and definitions;
• optimisation for reporting and analytics rather than transactional processing.

20. What is an outlier?

An outlier is an observation that differs substantially from the pattern of most other observations in a dataset.

Outliers can appear because of:

• data-entry or measurement errors;
• genuine rare events;
• changes in customer or system behaviour; or
• differences between underlying populations.

Common detection methods include box plots, the interquartile range (IQR), z-scores and visual exploration.

For example, under the common IQR rule, an observation may be flagged when it falls below Q1 − 1.5 × IQR or above Q3 + 1.5 × IQR.

An analyst should not automatically delete an outlier. The first step is to investigate whether it is an error or a meaningful observation.

21. How would you prepare raw data for analysis?

I would prepare the dataset systematically before starting the main analysis.

1. Understand the business question: determine which variables and level of detail are required.
2. Profile the data: check columns, data types, row counts, distributions and missing values.
3. Remove or resolve duplicates: identify the correct business key before deleting records.
4. Standardise formats: ensure dates, currencies, categories and units follow consistent rules.
5. Handle missing values and outliers: use a method appropriate to the business context.
6. Create useful derived fields: calculate variables needed for the analysis, such as customer tenure or monthly revenue.
7. Validate the final dataset: reconcile row counts and important totals before analysing it.

22. What questions should you ask before building a dashboard?

Before building a dashboard, I would clarify what decision it needs to support and who will use it.

• Who is the audience? Senior leaders, managers and analysts may need different levels of detail.
• What decisions should the dashboard support?
• Which KPIs are most important? Each metric should have a clear definition.
• How frequently should the data refresh? Real-time, daily, weekly or monthly?
• Which filters and dimensions are required? For example, region, product, channel or customer segment.
• What level of detail is needed? Should users be able to drill down?
• What are the trusted data sources?
• What action should users take when a KPI moves outside its expected range?

A good dashboard should answer a defined business question rather than display every available metric.

Statistics Interview Questions for Data Analysts

Statistics is a core part of data analysis because analysts need to interpret distributions, compare groups and distinguish meaningful patterns from random variation. These are some of the most useful statistics questions to prepare.

23. What is the difference between mean, median and mode?

For example, median income is often more informative than mean income when a small number of very high earners make the distribution strongly skewed.

24. What is standard deviation?

Standard deviation measures how spread out observations are around their mean.

A smaller standard deviation means values are clustered relatively close to the mean, while a larger standard deviation indicates greater variation.

For example, two sales teams may have the same average monthly sales but very different standard deviations. The team with the larger standard deviation has less consistent performance.

25. What is the difference between correlation and causation?

Correlation means two variables move together in some way. Causation means a change in one variable produces a change in another.

A correlation alone does not prove causation because the relationship may be caused by another variable, reverse causality or coincidence.

For example, higher advertising spend and higher sales may be correlated, but this alone does not establish how much additional sales were directly caused by advertising.

26. What is a normal distribution?

A normal distribution is a continuous probability distribution that is symmetric around its mean and has a bell-shaped curve.

For a perfectly normal distribution, the mean, median and mode are equal. Approximately 68% of observations fall within one standard deviation of the mean, about 95% within two and about 99.7% within three.

Not all business data follows a normal distribution, so analysts should examine the actual distribution before selecting statistical methods.

27. What is hypothesis testing?

Hypothesis testing is a statistical method used to evaluate whether the evidence in a sample is strong enough to support a conclusion about a wider population.

It usually begins with:

• Null hypothesis (H0): assumes there is no meaningful difference or effect.
• Alternative hypothesis (H1): proposes that a difference or effect exists.

The analyst then uses the sample data and an appropriate statistical test to determine how consistent the evidence is with the null hypothesis.

28. What is a p-value?

A p-value measures how compatible the observed result is with the null hypothesis, assuming the null hypothesis is true.

A small p-value indicates that the observed data would be relatively unusual under the null hypothesis. Analysts often compare it with a significance threshold such as 0.05.

However, a p-value does not tell you the probability that the null hypothesis is true, and statistical significance does not automatically mean the result is important from a business perspective.

29. What is the difference between Type I and Type II errors?

For example, in an A/B test, a Type I error could lead a company to believe a new page performs better when the apparent improvement happened by chance.

30. What is sampling bias?

Sampling bias occurs when the observations included in a sample are not representative of the population the analyst wants to understand.

For example, if a customer satisfaction survey is sent only to highly engaged customers, the results may overestimate satisfaction across the full customer base.

Analysts should examine how the data was collected, who was included or excluded and whether particular groups are disproportionately represented.

31. What is the difference between population and sample?

A population is the complete group you want to understand, while a sample is a subset of that population used for analysis.

For example, if a company wants to understand satisfaction among all 500,000 customers, those customers are the population. If it surveys 5,000 selected customers, those respondents form the sample.

A well-designed sample should be sufficiently representative of the population for the intended analysis.

32. What is confidence interval?

A confidence interval gives a range of plausible values for a population parameter based on sample data and the statistical procedure used.

For example, instead of reporting that average customer satisfaction is exactly 8.1, an analysis may report an estimated average along with a 95% confidence interval.

A narrower confidence interval indicates a more precise estimate, while a wider interval indicates greater uncertainty.

SQL Interview Questions for Data Analysts

SQL is one of the most important skills tested in a data analyst interview. Recruiters typically want to know whether you can retrieve data, combine tables, aggregate metrics, handle missing values, remove duplicates and solve business problems using SQL queries.

33. What is SQL?

SQL, or Structured Query Language, is used to interact with relational databases. Data analysts use SQL to retrieve, filter, join, aggregate and transform data stored in database tables.

Common SQL operations include:

• selecting required columns;
• filtering rows;
• joining multiple tables;
• grouping and aggregating data;
• creating calculated fields;
• ranking records; and
• identifying duplicates or missing values.

34. What is the difference between WHERE and HAVING in SQL?

WHERE filters individual rows before aggregation, while HAVING filters grouped results after aggregation.

SELECT department, COUNT(*) AS employee_count
FROM employees
WHERE status = 'Active'
GROUP BY department
HAVING COUNT(*) > 10;

In this example, WHERE keeps only active employees before the grouping takes place. HAVING then keeps only departments with more than 10 active employees.

35. What is GROUP BY used for?

GROUP BY combines rows that have the same values in one or more columns so aggregate functions such as COUNT(), SUM(), AVG(), MIN() and MAX() can be calculated for each group.

SELECT region, SUM(revenue) AS total_revenue
FROM sales
GROUP BY region;

This query calculates total revenue for each region.

36. What are the different types of SQL JOINs?

JOINs are used to combine rows from multiple tables based on a related column.

For data analyst interviews, INNER JOIN and LEFT JOIN are particularly common because analysts frequently need to combine transaction, customer, product or event tables.

37. What is the difference between INNER JOIN and LEFT JOIN?

An INNER JOIN returns only records that have a match in both tables. A LEFT JOIN keeps every row from the left table, even if no corresponding match exists in the right table.

For example, if you want a list of all customers, including those who have never placed an order, you would typically use a LEFT JOIN from the customer table to the orders table.

38. What is the difference between UNION and UNION ALL?

Both operators combine the results of multiple SELECT statements vertically.

• UNION: combines results and removes duplicate rows.
• UNION ALL: combines results and retains duplicates.

UNION ALL is generally faster because the database does not need to perform the additional duplicate-removal step.

39. What is a subquery?

A subquery is a query written inside another SQL query. It can be used to calculate an intermediate result that the outer query then uses.

SELECT employee_name, salary
FROM employees
WHERE salary > (
    SELECT AVG(salary)
    FROM employees
);

This query returns employees whose salary is greater than the overall average salary.

40. What is a Common Table Expression (CTE)?

A Common Table Expression, or CTE, is a temporary named result set created using the WITH clause. It exists only while the query is running.

WITH customer_sales AS (
    SELECT customer_id, SUM(amount) AS total_sales
    FROM orders
    GROUP BY customer_id
)

SELECT *
FROM customer_sales
WHERE total_sales > 50000;

CTEs are useful for making complex SQL queries easier to read, maintain and debug.

41. What are window functions in SQL?

Window functions perform calculations across a set of related rows while retaining the individual rows in the output.

Common window functions include:

• ROW_NUMBER()
• RANK()
• DENSE_RANK()
• LAG()
• LEAD()
• SUM() OVER()
• AVG() OVER()

They are widely used for ranking, running totals, period comparisons and identifying previous or next observations.

42. What is the difference between ROW_NUMBER(), RANK() and DENSE_RANK()?

For example, if three salaries are ₹100, ₹100 and ₹90, RANK() may produce 1, 1, 3, while DENSE_RANK() produces 1, 1, 2.

43. How would you find duplicate records in SQL?

You can identify duplicate values using GROUP BY and HAVING.

SELECT email, COUNT(*) AS record_count
FROM customers
GROUP BY email
HAVING COUNT(*) > 1;

This returns email addresses that appear more than once.

Before deleting duplicates, an analyst should identify the correct business key and determine which record should be retained.

44. How would you remove duplicate rows while keeping one record?

A common method is to assign a row number within each duplicate group and delete records where the row number is greater than one.

WITH duplicates AS (
    SELECT *,
           ROW_NUMBER() OVER (
               PARTITION BY email
               ORDER BY created_at
           ) AS rn
    FROM customers
)

SELECT *
FROM duplicates
WHERE rn = 1;

This example keeps the earliest record for each email. In a real database, the exact deletion syntax depends on the database engine and the business rule for choosing the correct record.

45. How does SQL handle NULL values?

NULL represents a missing or unknown value. It is not the same as zero, an empty string or FALSE.

To check for NULL values, use:

WHERE column_name IS NULL

or:

WHERE column_name IS NOT NULL

Functions such as COALESCE() can also be used to replace NULL with another value in query output.

SELECT COALESCE(phone_number, 'Not Available')
FROM customers;

46. What is CASE WHEN in SQL?

CASE WHEN adds conditional logic to a SQL query and can be used to create categories or calculated fields.

SELECT customer_id,
       total_spend,
       CASE
           WHEN total_spend >= 50000 THEN 'High Value'
           WHEN total_spend >= 20000 THEN 'Medium Value'
           ELSE 'Low Value'
       END AS customer_segment
FROM customers;

This query segments customers based on their total spending.

47. What is the difference between a primary key and a foreign key?

For example, customer_id may be the primary key in a customer table and appear as a foreign key in an orders table.

48. What is the difference between DELETE, TRUNCATE and DROP?

Exact transaction and rollback behaviour can vary across database systems, so in an interview it is better to explain the conceptual difference rather than make an absolute claim that applies to every SQL engine.

49. What is the difference between COUNT(*), COUNT(column) and COUNT(DISTINCT column)?

• COUNT(*) counts all rows.
• COUNT(column) counts rows where that column is not NULL.
• COUNT(DISTINCT column) counts unique non-NULL values in that column.

This distinction is important because using the wrong version can change the result of an analysis.

50. How would you find the second-highest salary in SQL?

One approach is to use DENSE_RANK().

WITH ranked_salaries AS (
    SELECT employee_name,
           salary,
           DENSE_RANK() OVER (
               ORDER BY salary DESC
           ) AS salary_rank
    FROM employees
)

SELECT employee_name, salary
FROM ranked_salaries
WHERE salary_rank = 2;

DENSE_RANK() is useful here because multiple employees can have the same salary.

51. How would you calculate a running total in SQL?

A running total can be calculated using a windowed SUM().

SELECT order_date,
       revenue,
       SUM(revenue) OVER (
           ORDER BY order_date
       ) AS cumulative_revenue
FROM daily_sales;

This returns cumulative revenue as the dates progress.

52. How would you calculate month-on-month growth in SQL?

You can first calculate monthly totals and then use LAG() to access the previous month’s value.

WITH monthly_sales AS (
    SELECT month,
           SUM(revenue) AS revenue
    FROM sales
    GROUP BY month
)

SELECT month,
       revenue,
       LAG(revenue) OVER (ORDER BY month) AS previous_month_revenue,
       100.0 * (
           revenue - LAG(revenue) OVER (ORDER BY month)
       ) / NULLIF(
           LAG(revenue) OVER (ORDER BY month), 0
       ) AS growth_percent
FROM monthly_sales;

This is a good example of how window functions are used in real analytical work rather than just in theoretical SQL questions.

53. How would you find customers who have never placed an order?

A common solution is a LEFT JOIN followed by a NULL check.

SELECT c.customer_id,
       c.customer_name
FROM customers c
LEFT JOIN orders o
    ON c.customer_id = o.customer_id
WHERE o.customer_id IS NULL;

The LEFT JOIN keeps all customers, while the NULL condition isolates customers with no matching order record.

54. What is the difference between a CTE and a subquery?

Both can be used to create intermediate query results, but they differ mainly in structure and readability.

• Subquery: written directly inside another query.
• CTE: defined separately at the beginning using the WITH clause.

For complex analytical queries, CTEs are often easier to read and debug. Performance differences depend on the database engine and query optimiser, so it is better not to assume that one is always faster than the other.

55. What is database normalisation?

Normalisation is the process of organising relational data into logically related tables to reduce unnecessary duplication and improve data integrity.

For example, rather than repeating customer details on every order record, customer information may be stored in a separate customer table and linked to the orders table through a key.

Denormalisation intentionally introduces some duplicated or pre-combined data to make certain analytical or reporting queries faster or simpler.

56. What is a SQL view?

A view is a named SQL query that presents data like a virtual table. It can simplify complex logic and provide users with a consistent representation of data without requiring them to repeatedly write the underlying query.

Views are commonly used to:

• simplify frequently used joins;
• standardise business logic;
• restrict access to particular columns or rows; and
• make reporting queries easier to maintain.

57. What is an index in SQL?

An index is a database structure that can speed up data retrieval by helping the database locate rows without scanning the entire table.

Indexes can improve read performance, especially on columns frequently used in filters or joins. However, they also consume storage and can add overhead to inserts, updates and deletes, so indexes should be designed carefully.

58. How would you approach a slow SQL query?

I would investigate the query systematically rather than immediately rewriting everything.

• Check whether unnecessary columns or rows are being retrieved.
• Review JOIN conditions and confirm they use the correct keys.
• Look for functions applied to heavily filtered columns that may prevent efficient index use.
• Check whether filtering can occur earlier in the query.
• Review whether large intermediate result sets are being created.
• Inspect the execution plan where access is available.
• Check whether appropriate indexes exist for frequently filtered or joined columns.

The best optimisation depends on the database engine, table size, data distribution and query workload.

59. Write a SQL query to find the top three products by revenue in each category.

This can be solved using aggregation followed by a ranking window function.

WITH product_revenue AS (
    SELECT category,
           product_id,
           SUM(revenue) AS total_revenue
    FROM sales
    GROUP BY category, product_id
),
ranked_products AS (
    SELECT category,
           product_id,
           total_revenue,
           DENSE_RANK() OVER (
               PARTITION BY category
               ORDER BY total_revenue DESC
           ) AS revenue_rank
    FROM product_revenue
)

SELECT category,
       product_id,
       total_revenue
FROM ranked_products
WHERE revenue_rank <= 3;

This type of question tests whether you can combine GROUP BY, CTEs and window functions in one analytical query.

60. How would you calculate the percentage contribution of each category to total sales?

SELECT category,
       SUM(revenue) AS category_revenue,
       100.0 * SUM(revenue)
       / SUM(SUM(revenue)) OVER () AS revenue_share_percent
FROM sales
GROUP BY category;

This query calculates category revenue and divides it by overall revenue using a window function.

Excel Interview Questions for Data Analysts

Excel remains one of the most widely used tools for data analysis, especially for quick calculations, reporting, reconciliation and exploratory work. Interviewers may test your knowledge of formulas, lookups, PivotTables, conditional aggregation and data-cleaning features.

61. What is the difference between VLOOKUP and XLOOKUP?

VLOOKUP searches for a value in the first column of a range and returns a value from another column in the same row. XLOOKUP is more flexible because the lookup and return ranges are specified separately.

For newer Excel versions, XLOOKUP is often easier to maintain, but VLOOKUP is still common in existing workbooks and interview questions.

62. What is the difference between SUMIF and SUMIFS?

SUMIF adds values based on one condition, while SUMIFS supports multiple conditions.

=SUMIF(A:A,"North",B:B)

This sums values in column B where column A equals “North”.

=SUMIFS(C:C,A:A,"North",B:B,"Electronics")

This sums column C where the region is North and the category is Electronics.

63. What is the difference between COUNT, COUNTA and COUNTIF?

• COUNT: counts cells containing numbers.
• COUNTA: counts non-empty cells.
• COUNTIF: counts cells that meet a specified condition.

For example, =COUNTIF(B:B,"Completed") counts how many cells in column B contain the value “Completed”.

64. What is an IF function in Excel?

The IF function returns one value when a condition is true and another when it is false.

=IF(B2>=80,"High","Low")

This can be used to classify values, create flags or apply basic business rules.

65. How would you find and remove duplicates in Excel?

Excel provides several ways to identify or remove duplicate records.

• Use Conditional Formatting to highlight duplicates.
• Use Remove Duplicates to eliminate repeated records.
• Use formulas such as COUNTIF() to flag duplicates before deleting them.
• Use Power Query when duplicate removal is part of a repeatable data-cleaning process.

Before removing records, verify which columns together define a true duplicate. Two rows may share one value, such as a customer name, without representing the same record.

66. What is Power Query in Excel?

Power Query is a data preparation and transformation tool available in Excel and Power BI. It allows analysts to connect to data sources, clean and reshape datasets, and repeat those transformation steps when the data is refreshed.

Common Power Query tasks include:

• removing unnecessary columns;
• changing data types;
• splitting or merging columns;
• handling null values;
• combining multiple files;
• merging datasets; and
• automating recurring data-cleaning steps.

67. What is conditional formatting?

Conditional formatting automatically changes the appearance of cells when specified rules are met.

Data analysts may use it to highlight:

• duplicate values;
• KPIs above or below a threshold;
• overdue dates;
• top or bottom performers;
• unusual values or potential errors.

It is useful for quick visual checks, but important analytical conclusions should still be based on the underlying data rather than colour alone.

68. What is the difference between absolute and relative cell references?

• Relative reference: A1 changes when the formula is copied.
• Absolute reference: $A$1 remains fixed.
• Mixed reference: $A1 or A$1 locks only the column or row.

For example, when applying a fixed tax rate stored in one cell across hundreds of rows, an absolute reference can prevent the tax-rate cell from shifting when the formula is copied.

69. How would you combine data from two Excel tables?

The appropriate method depends on the size and structure of the data.

• XLOOKUP or VLOOKUP: useful for bringing a value from one table into another based on a common key.
• INDEX and MATCH: another flexible lookup approach.
• Power Query Merge: better for repeatable joins or larger datasets.
• Append: useful when the tables have the same columns and need to be stacked vertically.

Before combining tables, confirm that the join key is reliable and understand whether it is one-to-one, one-to-many or many-to-many.

70. When would you use a PivotTable instead of formulas?

I would use a PivotTable when I need to quickly summarise a large dataset across multiple categories, such as revenue by month, region and product.

Formulas may be more suitable when the output requires a fixed customised layout or needs to feed other calculations. PivotTables are particularly useful for exploratory analysis because dimensions can be rearranged quickly without rewriting formulas.

Read Also: Power BI Interview Questions and Answers 2026

Python and Pandas Interview Questions for Data Analysts

Python is commonly used when analysis becomes difficult to manage through spreadsheets alone. For data analyst roles, interviewers are usually more interested in practical data manipulation using libraries such as Pandas and NumPy than in advanced software development concepts.

71. Why is Python useful for data analysis?

Python is useful because it can handle large datasets, automate repetitive tasks and support data cleaning, statistical analysis and visualisation within the same workflow.

Common libraries include:

• Pandas: tabular data manipulation and analysis.
• NumPy: numerical operations and arrays.
• Matplotlib: data visualisation.
• scikit-learn: machine learning and preprocessing when required.

72. What is a Pandas DataFrame?

A DataFrame is a two-dimensional labelled data structure in Pandas. It contains rows and columns and is similar conceptually to a spreadsheet or relational database table.

Analysts can use DataFrames to filter records, create new columns, aggregate values, join datasets, handle missing values and perform many other transformations.

73. What is the difference between loc and iloc in Pandas?

• loc selects data primarily by labels.
• iloc selects data by integer position.

df.loc[df["sales"] > 1000, ["customer", "sales"]]

df.iloc[0:5, 0:3]

The first example filters rows using a condition and selects named columns. The second selects rows and columns based on their numerical positions.

74. How do you handle missing values in Pandas?

Pandas provides several methods for identifying and handling missing values.

df.isna()

df.dropna()

df["age"] = df["age"].fillna(df["age"].median())

The method should depend on why the values are missing and how the treatment may affect the analysis. Automatically filling every missing value with the mean or median is not always appropriate.

75. How do you remove duplicates in Pandas?

You can identify duplicate records using duplicated() and remove them using drop_duplicates().

df[df.duplicated()]

df = df.drop_duplicates()

You can also identify duplicates based on selected columns:

df.drop_duplicates(
    subset=["customer_id"],
    keep="first"
)

As with SQL or Excel, confirm the correct business key before removing duplicate records.

76. What does groupby() do in Pandas?

groupby() divides data into groups and allows aggregate calculations to be performed for each group.

df.groupby("region")["revenue"].sum()

This calculates total revenue for every region.

Multiple aggregations can also be applied:

df.groupby("region")["revenue"].agg(
    ["sum", "mean", "count"]
)

77. What is the difference between merge() and concat() in Pandas?

merge() combines datasets based on matching keys, similar to SQL JOINs.

pd.merge(
    customers,
    orders,
    on="customer_id",
    how="left"
)

concat() combines DataFrames along rows or columns and is useful when datasets need to be stacked or placed side by side.

pd.concat(
    [january_sales, february_sales],
    ignore_index=True
)

78. How would you filter rows in a Pandas DataFrame?

Rows can be filtered using Boolean conditions.

df[df["revenue"] > 50000]

Multiple conditions can also be combined:

df[
    (df["region"] == "North")
    & (df["revenue"] > 50000)
]

79. What is the difference between apply() and vectorised operations in Pandas?

apply() can run a custom function across rows or columns. However, where possible, analysts generally prefer vectorised operations because they are usually faster and simpler for standard calculations.

df["profit"] = df["revenue"] - df["cost"]

This vectorised operation is generally preferable to applying a custom function row by row for the same calculation.

80. How would you identify the top five customers by revenue using Pandas?

top_customers = (
    df.groupby("customer_id")["revenue"]
      .sum()
      .sort_values(ascending=False)
      .head(5)
)

The process first groups transactions by customer, calculates total revenue, sorts the values in descending order and returns the top five customers.

81. How would you create a new calculated column in Pandas?

A new column can be created by assigning a calculation to a new column name.

df["profit_margin"] = (
    (df["revenue"] - df["cost"])
    / df["revenue"]
) * 100

Before performing such calculations, an analyst should also handle cases where the denominator may be zero or missing.

82. How would you read a CSV file in Pandas?

import pandas as pd

df = pd.read_csv("sales.csv")

After loading the file, an analyst would typically inspect the dataset using methods such as:

df.head()
df.info()
df.describe()
df.isna().sum()

These help quickly understand the structure, data types, basic distributions and missing-value patterns before deeper analysis begins.

Power BI and Tableau Interview Questions for Data Analysts

Business intelligence tools are commonly used to turn analysed data into dashboards and reports that stakeholders can use. Interviewers may test both your technical understanding and whether you can choose the right visual or metric for a business problem.

83. What is the difference between Power BI and Tableau?

Both Power BI and Tableau are business intelligence and data visualisation tools, but they differ in workflow, ecosystem and implementation.

The better tool depends on the organisation’s technology stack, reporting needs, licensing and user requirements.

84. What is DAX in Power BI?

DAX, or Data Analysis Expressions, is the formula language used in Power BI, Power Pivot and Analysis Services for calculations such as measures, calculated columns and time-based analysis.

Examples include calculating total sales, profit margin, year-to-date revenue or percentage contribution.

Total Sales = SUM(Sales[Revenue])

A key interview point is that DAX calculations are affected by filter context, which determines which rows are included in a measure at a given point in a report.

85. What is the difference between a calculated column and a measure in Power BI?

For example, a calculated column could classify each transaction as high or low value, while a measure could calculate total revenue for whichever region the user selects.

86. What is a relationship in Power BI?

A relationship connects tables through common fields so that filters and calculations can work across the data model.

For example, a Sales table may contain product_id, while a Product table stores the product name and category. A relationship allows a report to analyse sales by those product attributes.

Interviewers may also ask about cardinality such as:

• one-to-one;
• one-to-many;
• many-to-one; and
• many-to-many.

For analytical models, one-to-many relationships between dimension and fact tables are common.

87. What is a star schema?

A star schema is a data modelling structure where a central fact table is connected to surrounding dimension tables.

For example:

• Fact table: Sales transactions
• Dimension table: Customers
• Dimension table: Products
• Dimension table: Date
• Dimension table: Region

Star schemas are commonly used in BI models because they make relationships and calculations easier to understand and can improve reporting performance.

88. What makes a good dashboard?

A good dashboard helps a user understand performance and take action without unnecessary complexity.

• It focuses on a small number of important KPIs.
• Metrics have clear and consistent definitions.
• Visuals are chosen according to the question being answered.
• Important information appears prominently.
• Filters are useful but not excessive.
• Colours and labels are consistent.
• The dashboard supports comparison over time or against a target where relevant.

The objective is not to fit as many charts as possible on one screen. Every visual should help answer a business question.

89. Which chart would you use for different types of analysis?

The best chart depends on what comparison the viewer needs to make. Visual clarity should take priority over decorative complexity.

90. How would you improve a slow Power BI dashboard?

I would first identify whether the problem comes from the data model, calculations, data volume or report design.

• Remove unnecessary columns and rows before loading data.
• Use an efficient star-schema model where appropriate.
• Avoid unnecessary many-to-many relationships.
• Review expensive DAX measures.
• Reduce the number of visuals on a single page.
• Avoid loading highly granular data when it is not required.
• Move suitable transformations upstream or into Power Query.
• Use performance analysis tools to identify slow visuals or calculations.

The exact solution depends on where the performance bottleneck occurs.

91. What are dimensions and measures in Tableau?

In Tableau, dimensions are generally fields used to categorise, group or segment data, while measures are typically numerical fields that can be aggregated.

For example:

• Dimensions: region, customer name, product category
• Measures: revenue, quantity, profit

An analyst may use the Region dimension to break total Revenue into separate groups.

92. What is the difference between a filter and a parameter in Tableau?

A filter controls which data is included in a view. A parameter is a user-controlled value that can be used in calculations, reference lines, filters or other interactive logic.

For example, a parameter could allow a user to choose between Revenue, Profit and Quantity as the metric displayed in a chart.

Read Also: Power BI Interview Questions and Answers 2026

Scenario-Based Data Analyst Interview Questions

Scenario-based questions test how you approach an unclear business problem, validate the data and communicate your reasoning. There may not be one perfect answer. Interviewers usually look for a structured approach and sensible assumptions.

93. Sales dropped by 20% last month. How would you investigate?

I would avoid immediately assuming the cause and investigate the drop systematically.

1. Validate the metric: confirm that the 20% decline is real and not caused by incomplete data or a reporting change.
2. Compare periods: check month-on-month and year-on-year trends to understand whether seasonality may be involved.
3. Segment the decline: break sales down by region, product, customer segment, channel and salesperson.
4. Separate volume and price: determine whether the decline came from fewer transactions, lower prices or both.
5. Check the funnel: analyse traffic, leads, conversions, order values and cancellations where relevant.
6. Look for operational changes: investigate stock-outs, campaign changes, pricing, distribution issues or product availability.
7. Summarise the drivers: quantify which factors explain the largest part of the decline.

The final output should explain both what changed and which factors contributed most to the change.

94. Website traffic increased but conversions decreased. What would you check?

I would first check whether the additional traffic came from the same type of users as before.

• Traffic source and campaign mix
• New versus returning visitors
• Device and browser
• Landing pages
• Geography
• Page-load or technical issues
• Conversion funnel drop-off
• Offer, pricing or product availability

For example, traffic may rise because of a broad awareness campaign that brings many low-intent visitors. In that case, total visits increase while the overall conversion rate falls.

95. A stakeholder says your dashboard number is wrong. What would you do?

I would first clarify which metric and period they believe is incorrect instead of defending the dashboard immediately.

1. Confirm the exact metric definition being compared.
2. Check whether both parties are using the same date range, filters and source.
3. Trace the calculation back to the underlying data.
4. Review joins, filters and transformation logic.
5. Reconcile totals against the source system where possible.
6. Document the finding and correct the dashboard if an error exists.

Sometimes the problem is not a calculation error but a difference in metric definitions. For example, one team may count orders created during a month while another counts orders completed during the month.

96. Two reports show different revenue numbers. How would you reconcile them?

I would compare the reports systematically across the following areas:

• source systems;
• date ranges and time zones;
• currency conversion;
• gross versus net revenue;
• cancelled or refunded orders;
• filters and exclusions;
• join logic;
• data refresh timing; and
• metric definitions.

I would then reconcile the reports at a lower level, such as daily or transaction-level totals, until the source of the difference becomes clear.

97. How would you measure the success of a new product feature?

I would begin by asking what business or user problem the feature is intended to solve.

Possible metrics could include:

• Adoption: percentage of eligible users who use the feature.
• Engagement: frequency or depth of feature usage.
• Conversion: whether usage improves the desired funnel outcome.
• Retention: whether users who adopt the feature remain more active.
• Efficiency: whether the feature reduces time or effort.
• Guardrail metrics: whether the feature negatively affects other important outcomes.

Where possible, I would compare the outcome against a baseline, control group or pre-launch period rather than looking only at raw usage.

98. How would you analyse customer churn?

I would first define exactly what counts as churn. The definition could differ between a subscription business, e-commerce platform and mobile application.

1. Define the churn event and observation period.
2. Calculate the overall churn rate.
3. Segment churn by customer tenure, plan, acquisition channel, geography or product usage.
4. Compare churned and retained customers across behavioural variables.
5. Analyse changes before the churn event, such as declining usage or repeated support issues.
6. Identify actionable drivers rather than correlations alone.
7. Track whether retention actions improve the outcome.

99. A dataset shows a sudden spike in one metric. What would you do?

I would investigate whether the spike represents a real business event or a data-quality issue.

• Check whether the spike appears in the source data.
• Compare it with related metrics.
• Review recent tracking, pipeline or schema changes.
• Break the spike down by source, geography, product or customer segment.
• Check whether duplicates or delayed records were loaded.
• Compare against historical events and seasonality.

I would not remove the unusual observation simply because it looks like an outlier. First I would determine whether it represents an error or an important business event.

100. Your manager asks for a report by the end of the day, but the data has quality issues. What would you do?

I would explain the specific data-quality issue, estimate how it may affect the result and prioritise the checks that are most important to the business decision.

If a reliable partial analysis can be delivered, I would clearly label its limitations and avoid presenting uncertain numbers as final. For a critical issue that could materially change the conclusion, I would escalate it rather than knowingly publish an unreliable result.

The goal is to balance speed with analytical integrity and make the uncertainty visible to the stakeholder.

101. How would you prioritise multiple data requests from stakeholders?

I would prioritise requests based on factors such as:

• business impact;
• decision deadline;
• number of teams affected;
• effort required;
• availability and quality of the required data; and
• whether another analysis depends on the result.

I would also clarify the actual decision behind each request. Sometimes a stakeholder asks for a detailed analysis when a much smaller calculation can answer the immediate business question.

102. How would you explain a technical finding to a non-technical stakeholder?

I would focus on the business meaning rather than the technical process.

A useful structure is:

1. What happened?
2. Why does it matter?
3. What is likely driving it?
4. What action should we consider?
5. How confident are we in the conclusion?

For example, instead of explaining a regression coefficient in technical terms, I might say: “Customers who contact support three or more times in a month are much more likely to cancel, so this group may be worth targeting with proactive support.”

Behavioural Data Analyst Interview Questions

Behavioural questions test how you work with people, manage ambiguity, handle mistakes and communicate analytical findings. Use examples from internships, college projects, freelance work or previous roles where possible.

103. Tell me about a time when your analysis changed a business decision.

Structure your answer around the problem, analysis and outcome. Explain what decision was being considered, what data you used, what insight you found and what changed because of your recommendation.

For example: “The marketing team planned to increase spending on a campaign because it was generating high traffic. I analysed conversions by source and found that the campaign had one of the lowest conversion rates. The team shifted part of the budget to a higher-converting channel, improving the overall return on campaign spend.”

104. Tell me about a mistake you made while analysing data.

Choose a genuine mistake and focus on how you detected, corrected and prevented it from happening again.

A strong answer could involve using the wrong join type, applying an incorrect filter, misunderstanding a metric definition or overlooking duplicate records.

Explain the control you introduced afterwards, such as row-count checks, reconciliation against source totals, peer review or documenting metric definitions.

105. How do you handle unclear requirements from a stakeholder?

I would clarify the decision the stakeholder is trying to make before beginning the analysis.

• What business question needs to be answered?
• Who will use the analysis?
• Which metric matters most?
• What time period and level of detail are required?
• What action will be taken based on the result?

If the requirement is still ambiguous, I would share a small sample or initial analysis early so the stakeholder can confirm that the direction is correct before significant work is completed.

106. How do you handle disagreement with a stakeholder about your findings?

I would first understand whether the disagreement is about the data, metric definition, assumptions or interpretation.

I would walk through the methodology and evidence, check whether the stakeholder has additional context, and revise the analysis if new information changes the conclusion.

The objective is not to defend the original answer at all costs. It is to reach the most accurate interpretation of the data.

107. How do you ensure accuracy in your analysis?

I use checks at different stages of the analysis rather than relying only on the final result.

• Validate data types, row counts and missing values.
• Check for duplicate records.
• Reconcile important totals against source systems.
• Validate joins by checking row-count changes.
• Test calculations on small samples manually.
• Compare results with historical trends or expected ranges.
• Document assumptions and metric definitions.
• Use peer review for high-impact analysis where possible.

108. How do you manage tight deadlines?

I first identify the minimum analysis required to support the decision and prioritise the highest-risk validation checks.

If the full analysis cannot be completed reliably within the deadline, I would communicate what can be delivered, what remains uncertain and what additional work is required. I would avoid sacrificing critical validation simply to produce a complete-looking report.

109. How do you keep your data analysis skills up to date?

I combine practical work with structured learning. This may include practising SQL problems, building small analytics projects, reviewing new Excel or BI features, improving Python skills and studying how companies define and measure business metrics.

For interview preparation, applying a tool to a real dataset is generally more useful than only memorising definitions.

110. Why do you want to work as a data analyst?

Your answer should connect your skills with the nature of the role rather than giving a generic response about liking data.

For example: “I enjoy breaking down business problems and using data to understand what is actually happening. Data analytics combines the technical work I enjoy, such as SQL and Excel, with business problem-solving and communication. I also like that the output of the analysis can directly influence decisions.”

How to Prepare for a Data Analyst Interview

Data analyst interviews can vary considerably by company, but preparation should cover both technical skills and business problem-solving.

1. Practise SQL with real problems

Do not limit preparation to SQL definitions. Practise writing queries involving:

• JOINs;
• GROUP BY and HAVING;
• CTEs and subqueries;
• window functions;
• NULL handling;
• duplicates;
• date-based calculations; and
• ranking and aggregation problems.

2. Revise Excel fundamentals

Be comfortable with PivotTables, XLOOKUP or VLOOKUP, SUMIFS, COUNTIFS, IF functions, conditional formatting and basic data cleaning. Power Query is also useful for roles that involve recurring reporting or spreadsheet-based workflows.

3. Understand basic statistics

Prepare concepts such as mean versus median, standard deviation, correlation, sampling, confidence intervals, hypothesis testing and p-values. Focus on explaining what these concepts mean in a business context rather than memorising formulas alone.

4. Prepare two or three projects thoroughly

For each project, be ready to explain:

• the business problem;
• the data source;
• data-quality issues;
• tools used;
• your analytical approach;
• the main insight; and
• the final outcome or recommendation.

Interviewers may ask follow-up questions about why you selected a particular method, so understand the reasoning behind your choices.

5. Practise business case questions

Prepare to investigate situations such as falling sales, declining conversions, increasing churn or inconsistent dashboard numbers. A structured approach is more valuable than jumping immediately to a conclusion.

6. Know the company’s key metrics

Before the interview, understand how the company makes money and which metrics may matter to its business model. For example, an e-commerce company may focus on conversion rate, average order value and repeat purchases, while a subscription business may focus on retention, churn and recurring revenue.

7. Explain your reasoning while solving questions

If an interviewer gives you a SQL problem, estimation question or business case, explain your assumptions and reasoning. Interviewers often evaluate how you approach an unfamiliar problem, not only whether you reach the final answer immediately.

Conclusion

Preparing for a data analyst interview requires more than memorising definitions. Candidates should be comfortable with SQL, Excel, statistics, data cleaning, dashboards and business problem-solving, while also being able to explain their findings clearly.

Freshers should focus on strong fundamentals and be ready to explain academic or personal projects in detail. Experienced candidates should additionally prepare for questions on stakeholder management, metric definitions, data quality, dashboard design and scenario-based analysis.

The strongest interview answers usually explain not only what you would do, but also why you would choose that approach and how you would validate the result.

More Resources: Job vacancies in Kochi | Job vacancies in Kuwait | Short resignation letter | Technical interview questions and answers

FAQs

What questions are asked in a data analyst interview?

Data analyst interviews commonly include questions on SQL, Excel, statistics, data cleaning, dashboards, Python or R, business metrics and analytical problem-solving. Experienced candidates may also face scenario-based and stakeholder-management questions

Is SQL important for a data analyst interview?

Yes. SQL is one of the most commonly tested technical skills for data analyst roles. Candidates should prepare JOINs, GROUP BY, WHERE and HAVING, CTEs, subqueries, window functions, NULL handling and analytical SQL problems

What Excel skills should a data analyst know?

Useful Excel skills include PivotTables, XLOOKUP or VLOOKUP, SUMIFS, COUNTIFS, IF functions, conditional formatting, data cleaning and Power Query. The exact level required depends on the role

Do data analysts need Python?

Python is useful for data cleaning, automation and analysis, especially when datasets become difficult to manage in spreadsheets. However, some entry-level analyst roles rely more heavily on SQL, Excel and BI tools, so requirements vary by employer.

How should freshers prepare for a data analyst interview?

Freshers should strengthen SQL, Excel and basic statistics, practise data-cleaning problems, build at least two practical analytics projects and prepare to explain the business question, methodology and insights from each project.

How do you answer scenario-based data analyst interview questions?

Start by clarifying the business question, validate the data, break the problem into relevant segments, identify possible drivers and explain how you would test them. State assumptions clearly and avoid jumping to conclusions before checking the evidence.