CareerGuide 7 of 36
Career guide · Guide 7 of 36
Top 50 Data Analyst Interview Questions and Answers
Top 50 data analyst interview questions with simple answers for freshers and experienced: SQL, Excel, statistics, A/B testing, metrics and case study questions.
On this page
These 50 questions cover what data analyst interviews ask, for freshers and experienced candidates: the role and process, SQL, Excel, statistics and A/B testing, metrics, and the case study questions that test how you think. Each answer is written in plain words and gets straight to the point. For BI tool questions, see the Top 50 Power BI interview questions.
Question list
Show all 50 questions
- What does a data analyst do?
- What are the steps of a data analysis project?
- What is data cleaning, and why is it important?
- How do you handle missing values?
- What is the difference between data analysis and data analytics?
- What are the types of data analytics?
- What is the difference between structured and unstructured data?
- What is the difference between mean, median and mode, and when do you use each?
- What is an outlier, and how do you deal with one?
- What is the difference between a KPI and a metric?
- Which tools does a data analyst use?
- What is a pivot table?
- What is VLOOKUP, and what is better about XLOOKUP?
- What is data visualisation, and how do you choose a chart?
- How do you explain findings to non-technical stakeholders?
- What is the difference between WHERE and HAVING?
- Explain the types of SQL joins.
- How do you find duplicate rows in SQL?
- What are window functions, and when do you use them?
- How would you find the second-highest salary in SQL?
- How do you calculate month-over-month growth in SQL?
- What is a CTE, and why use it?
- What is cohort analysis?
- What is funnel analysis?
- What is A/B testing?
- What is a p-value?
- What is the difference between correlation and causation?
- What is standard deviation?
- What is a confidence interval?
- What is the difference between COUNT, COUNTA, COUNTIF and COUNTIFS in Excel?
- How would you clean a messy dataset in Excel?
- How do you measure whether a dashboard is useful?
- What is the difference between a data analyst, data scientist and data engineer?
- What is the difference between a data analyst and a business analyst?
- What is normalisation in databases, and why should an analyst care?
- Case: sales dropped 20% last month. How do you investigate?
- Case: how would you measure the success of a new feature?
- How do you decide the sample size for an A/B test?
- What are common mistakes in A/B testing?
- What is Simpson’s paradox?
- How do you prioritise many requests from different stakeholders?
- Tell me about an analysis that changed a decision.
- How do you make sure your numbers are correct?
- What is RFM analysis?
- How do you calculate customer churn and retention?
- How would you build a sales forecast as an analyst?
- How do you handle a stakeholder who disagrees with your analysis?
- How do you present an analysis in a five-minute meeting?
- What is the difference between a metric definition and a dashboard?
- What would you do in your first 90 days as a data analyst?
Beginner data analyst interview questions (freshers)
1. What does a data analyst do?
A data analyst turns data into answers for the business. They collect and clean data, explore it, calculate metrics, find trends and explain what they mean, usually through reports, dashboards and short write-ups. Typical questions are “why did sales drop last month?” or “which customers are most likely to leave?”. The main tools are SQL, Excel or Google Sheets, a BI tool such as Power BI or Tableau, and often Python, plus basic statistics and clear communication.
2. What are the steps of a data analysis project?
First, define the question and how the answer will be used. Second, collect the data needed. Third, clean it: fix types, handle missing values and duplicates, and check it against known totals. Fourth, explore it with summaries and charts. Fifth, analyse: calculate metrics, compare groups, test ideas. Sixth, communicate the findings with a clear recommendation. Finally, check what happened after the decision and improve the analysis.
Most of the time goes into steps two and three. Interviewers like candidates who mention checking the data before trusting it, and who end with a clear recommendation rather than just charts.
3. What is data cleaning, and why is it important?
Data cleaning means finding and fixing problems in data before analysing it: duplicates, missing values, wrong types, inconsistent spellings or codes, impossible values and outliers caused by errors. It matters because conclusions are only as good as the data; a duplicated order table can double revenue in a report. Cleaning usually takes a large share of an analyst’s time, so document every rule you apply so others can repeat it.
4. How do you handle missing values?
First find out why they are missing: not collected, not applicable, or lost by an error. Then choose: leave them and report the gap, remove rows if few are missing and they are random, fill them with a sensible value (zero only when missing truly means zero, or a median for a numeric field), or fill them from another source. Never fill silently; state how many values were missing and what you did, because it can change the result.
5. What is the difference between data analysis and data analytics?
The terms are often used as synonyms. When people separate them, data analysis means examining a specific dataset to answer a question, while data analytics is the broader practice that includes collecting, storing, processing and analysing data, building dashboards and making predictions. In interviews, give a short answer and focus on what you do: turn data into decisions.
Example answer: “Analysis is answering one question with data, such as why returns rose in March. Analytics is the whole practice and set of tools that lets a company do that regularly, from data pipelines to dashboards.”
6. What are the types of data analytics?
Descriptive analytics explains what happened, such as last quarter’s sales by region. Diagnostic analytics explains why it happened, such as which product drove the drop. Predictive analytics estimates what will happen, such as next month’s demand. Prescriptive analytics recommends what to do, such as how much stock to order. Most analyst work is descriptive and diagnostic, with some simple prediction.
Example for one business: a dashboard shows sales fell 10% (descriptive); a breakdown shows the fall came from one region after a price change (diagnostic); a model estimates next month’s sales (predictive); and a recommendation suggests reversing the price change there (prescriptive).
7. What is the difference between structured and unstructured data?
Structured data fits neatly into rows and columns with a fixed schema, like a sales table in a database, and is easy to query with SQL. Unstructured data has no fixed format, like emails, PDFs, images or call recordings. Semi-structured data, such as JSON or log files, has some structure through keys or tags. Analysts mostly work with structured data, but increasingly extract structured fields from the others.
Example: customer reviews are unstructured text, but you can turn them into structured data by tagging each review with a rating, topic and sentiment.
8. What is the difference between mean, median and mode, and when do you use each?
The mean is the average: the sum divided by the count. The median is the middle value when sorted. The mode is the most frequent value. Use the mean for symmetric data without extreme values; use the median for skewed data such as salaries or house prices, because a few very large values pull the mean up. Use the mode for categories, such as the most common product size.
Example: salaries of 30, 32, 35, 38 and 200 thousand have a mean of 67 thousand but a median of 35 thousand; the median describes a typical employee much better.
Learn more: Descriptive Statistics: Mean, Median, Spread and Outliers
9. What is an outlier, and how do you deal with one?
An outlier is a value far from the rest of the data. Find them with box plots, the interquartile range rule (below Q1 - 1.5 x IQR or above Q3 + 1.5 x IQR) or z-scores. Then investigate: if it is an error, such as a price typed with extra zeros, fix or remove it; if it is real, such as one very large customer, keep it but consider reporting results with and without it, or using the median.
Learn more: Descriptive Statistics: Mean, Median, Spread and Outliers
10. What is the difference between a KPI and a metric?
A metric is any number you measure, such as page views or number of orders. A KPI (key performance indicator) is a metric chosen because it shows progress towards an important business goal, usually with a target, such as monthly recurring revenue or on-time delivery rate. All KPIs are metrics, but only a few metrics are KPIs. Good KPIs are clearly defined, owned by someone and linked to a decision.
11. Which tools does a data analyst use?
The usual toolkit:
- SQL to get and combine data from databases.
- Excel or Google Sheets for quick analysis, pivot tables and sharing.
- A BI tool such as Power BI, Tableau or Looker Studio for dashboards.
- Python with pandas, or R, for larger or repeatable analysis and statistics. Analysts also use a version control or documentation tool and their company’s data warehouse. Strong SQL and Excel are the most commonly tested skills.
12. What is a pivot table?
A pivot table summarises a large table quickly by grouping and aggregating it. You drag fields into rows, columns, values and filters, for example region in rows, month in columns and sum of sales in values, and Excel calculates the totals for each combination. You can change the summary instantly, show percentages, and drill into the detail. It is one of the most useful tools for fast exploration and is asked about in most analyst interviews.
Learn more: PIVOT, UNPIVOT, GROUPING SETS, ROLLUP and CUBE
13. What is VLOOKUP, and what is better about XLOOKUP?
VLOOKUP looks up a value in the first column of a range and returns a value from a column to its right, using a column number. XLOOKUP is its modern replacement: it can look left or right, uses a return range instead of a column number (so inserting columns does not break it), does an exact match by default, and has a built-in value for when nothing is found. INDEX with MATCH is the classic alternative that works in older versions.
14. What is data visualisation, and how do you choose a chart?
Data visualisation shows data as charts so patterns are easy to see. Choose by the question: a line chart for change over time, a bar chart to compare categories, a scatter plot for the relationship between two numbers, a histogram for a distribution, and a single number with a comparison for a KPI. Keep charts simple: clear titles, labelled axes, no 3D effects, and colour used only to highlight what matters.
15. How do you explain findings to non-technical stakeholders?
Lead with the answer and the recommended action, then give two or three supporting facts, and keep the details for questions. Use the business’s words, not technical ones, and show one clear chart rather than many. Say how confident you are and what the main limitations are. For example: “Churn rose because of the price change in March; reversing it for small customers would likely win back about half of them.”
SQL, Excel and statistics questions for data analysts
16. What is the difference between WHERE and HAVING?
WHERE filters rows before they are grouped, so it cannot use aggregate results. HAVING filters groups after GROUP BY, so it can use aggregates such as SUM or COUNT. For example, to find cities with more than 100 orders in 2026: filter the year with WHERE, group by city, then keep groups with HAVING COUNT(*) > 100. Filtering early with WHERE is also faster, because fewer rows are grouped.
Example:
SELECT city, COUNT(*) AS orders
FROM orders
WHERE order_date >= '2026-01-01'
GROUP BY city
HAVING COUNT(*) > 100;Learn more: SQL Aggregations: GROUP BY, HAVING and Conditional Aggregation
17. Explain the types of SQL joins.
An INNER JOIN keeps only rows that match in both tables. A LEFT JOIN keeps every row from the left table and fills unmatched right-side columns with NULL; it is the most used join in analysis. A RIGHT JOIN is the mirror image. A FULL OUTER JOIN keeps all rows from both sides. A CROSS JOIN pairs every row with every row. Always check the row count after a join, because duplicate keys multiply rows. See SQL joins.
18. How do you find duplicate rows in SQL?
Group by the columns that should be unique and keep groups with more than one row: SELECT email, COUNT() FROM customers GROUP BY email HAVING COUNT() > 1. To see or remove the extra copies, number the rows within each group with ROW_NUMBER() OVER (PARTITION BY email ORDER BY created_at) and keep row 1. Decide first which copy is the right one to keep, such as the most recent.
Example that keeps the latest row per email:
SELECT *
FROM (
SELECT c.*, ROW_NUMBER() OVER (PARTITION BY email ORDER BY created_at DESC) AS rn
FROM customers c
) t
WHERE rn = 1;Learn more: Top-N per Group, Deduplication and SCD Queries
19. What are window functions, and when do you use them?
Window functions calculate a value for each row using a set of related rows, without collapsing them into one row like GROUP BY. Analysts use them for rankings (ROW_NUMBER, RANK), running totals (SUM OVER ORDER BY date), comparing with the previous period (LAG), and percent of total. For example, each order’s share of its customer’s total spend. See window functions.
Example, running total of revenue per customer:
SELECT customer_id, order_date, amount,
SUM(amount) OVER (PARTITION BY customer_id ORDER BY order_date) AS running_total
FROM orders;Every order row stays in the result, and each one shows the customer’s total spend up to that date.
20. How would you find the second-highest salary in SQL?
Use DENSE_RANK so ties are handled: SELECT salary FROM (SELECT salary, DENSE_RANK() OVER (ORDER BY salary DESC) AS r FROM employees) t WHERE r = 2. A classic alternative is SELECT MAX(salary) FROM employees WHERE salary < (SELECT MAX(salary) FROM employees). Mention what should happen if two people share the top salary, and if there is no second value, because interviewers check that you think about edge cases.
To get it per department, add PARTITION BY department to the window function and filter on the rank in the same way.
Learn more: Top-N per Group, Deduplication and SCD Queries
21. How do you calculate month-over-month growth in SQL?
Aggregate to one row per month first, then compare each month with the previous one using LAG:
WITH m AS (
SELECT DATE_TRUNC('month', order_date) AS month, SUM(amount) AS revenue
FROM orders
GROUP BY 1
)
SELECT month, revenue,
revenue / LAG(revenue) OVER (ORDER BY month) - 1 AS growth
FROM m;Handle the first month (no previous value) and months with zero revenue, and make sure every month appears even with no orders.
Learn more: Time-Series SQL: Date Buckets, Calendars, YoY, MoM and Rolling Metrics
22. What is a CTE, and why use it?
A common table expression (CTE) is a named temporary result defined with WITH at the start of a query. It lets you build a query in readable steps, such as first calculate each customer’s total, then compare it with the average, instead of nesting subqueries. CTEs make queries easier to read, test and change. Some databases also support recursive CTEs for hierarchies. See CTEs and subqueries.
Example:
WITH totals AS (
SELECT customer_id, SUM(amount) AS spend
FROM orders
GROUP BY customer_id
)
SELECT customer_id, spend
FROM totals
WHERE spend > (SELECT AVG(spend) FROM totals);This finds customers who spend more than the average customer, in two readable steps.
23. What is cohort analysis?
Cohort analysis groups users by a shared starting event, usually the month they signed up or first bought, and tracks how each group behaves over time, such as the share still active after 1, 2 and 3 months. It shows whether retention is improving for newer customers, which a single overall number can hide. The result is usually shown as a triangle-shaped table or a set of lines. See retention and cohort analysis in SQL.
24. What is funnel analysis?
Funnel analysis measures how many users complete each step of a process, such as visit, add to cart, checkout and payment, and where they drop off. You calculate the conversion rate between each pair of steps and overall. It shows where to focus improvements: a big drop at payment suggests a problem with the payment page. Define steps and time windows clearly, and count users rather than events.
Example: 10,000 visitors, 2,000 add to cart (20%), 800 reach checkout (40%), 600 pay (75%). The biggest drop is from visit to cart, so that is where to look first.
Learn more: Product Analytics SQL: Retention, Cohorts, Funnels and Attribution
25. What is A/B testing?
A/B testing compares two versions of something, such as two checkout page designs, by randomly showing each version to a different group of users and measuring a chosen metric, such as conversion rate. Random assignment means differences in results are caused by the change, not by who saw it. Decide the metric, sample size and test length in advance, then use a statistical test to check whether the difference is likely to be real.
Learn more: A/B Testing: Design, Sample Size and Analysis
26. What is a p-value?
A p-value is the probability of seeing a result at least as extreme as yours if there were really no effect (the null hypothesis is true). A small p-value, commonly below 0.05, means the result would be unusual by chance alone, so you reject the null hypothesis. It does not tell you the size or business importance of the effect, and it is not the probability that your hypothesis is true. Report it with the effect size and a confidence interval.
Learn more: Hypothesis Testing and P-Values Explained Simply
27. What is the difference between correlation and causation?
Correlation means two variables move together; causation means one actually causes the other. Ice-cream sales and drowning deaths are correlated because both rise in summer, not because one causes the other. A hidden third factor, chance or reverse causation can create correlation. To show causation you need a controlled experiment, such as an A/B test, or careful methods that rule out other explanations.
In interviews, when you find a correlation, say how you would test whether it is causal, for example with an A/B test or by comparing similar groups where only one factor differs.
28. What is standard deviation?
Standard deviation measures how spread out values are around the mean. A small standard deviation means most values are close to the average; a large one means they vary widely. For roughly normal data, about 68% of values fall within one standard deviation of the mean and about 95% within two. Analysts use it to judge variability, spot unusual values and compare the consistency of groups.
Example: two stores both average 100 orders a day, but one ranges from 95 to 105 and the other from 40 to 160. Same mean, very different standard deviation, and very different staffing needs.
Learn more: Descriptive Statistics: Mean, Median, Spread and Outliers
29. What is a confidence interval?
A confidence interval gives a range of plausible values for something you estimate from a sample, such as “average order value is 52 dollars, with a 95% confidence interval of 49 to 55”. If you repeated the sampling many times, about 95% of such intervals would contain the true value. Narrow intervals mean precise estimates; larger samples give narrower intervals. Reporting the interval is more honest than reporting a single number.
Learn more: Hypothesis Testing and P-Values Explained Simply
30. What is the difference between COUNT, COUNTA, COUNTIF and COUNTIFS in Excel?
COUNT counts cells that contain numbers. COUNTA counts cells that are not empty, of any type. COUNTIF counts cells that meet one condition, such as COUNTIF(B:B, “Delhi”). COUNTIFS counts rows that meet several conditions at once, such as city is Delhi and amount is over 1,000. The same pattern applies to SUMIF, SUMIFS and AVERAGEIFS, which are used constantly in analysis.
Example: =COUNTIFS(B:B,“Delhi”,C:C,“>1000”) counts orders from Delhi worth more than 1,000.
31. How would you clean a messy dataset in Excel?
Keep the raw data untouched on its own sheet. Remove duplicates, trim extra spaces with TRIM, fix text case with PROPER or UPPER, split or combine columns, convert text numbers and dates to real types, and standardise categories with Find and Replace or a lookup table. Use filters to find blanks and odd values. For repeatable cleaning, use Power Query, which records every step so the same cleaning runs on next month’s file.
32. How do you measure whether a dashboard is useful?
Check whether it changes decisions: ask the intended users what they look at and what they did because of it. Look at usage data, such as views per week and which pages are used. Check accuracy against the source, and loading speed. A useful dashboard answers a few clear questions, is trusted and is used regularly; if not, simplify it or retire it.
A simple question to ask users: “If this dashboard disappeared tomorrow, what would you miss?” If the answer is “nothing”, it needs rethinking.
33. What is the difference between a data analyst, data scientist and data engineer?
A data engineer builds the systems that collect and prepare data. A data analyst uses that data to answer business questions with SQL, spreadsheets and dashboards. A data scientist builds statistical and machine learning models to predict or explain outcomes. They overlap, especially in small companies. See data engineer vs data scientist vs data analyst.
A simple way to say it: the data engineer makes data available and reliable, the analyst explains what happened and why, and the data scientist predicts what will happen. They work best when they understand each other’s needs.
34. What is the difference between a data analyst and a business analyst?
A data analyst focuses on data: querying, cleaning, analysing and reporting on it to answer questions. A business analyst focuses on business processes and requirements: understanding how the business works, documenting what a system or change must do, and working between business teams and developers. Business analysts use data too, but less technically. Many roles mix both, so read the job description.
Example: for a new returns process, the business analyst writes down how returns should work and what the system must do; the data analyst measures how many returns happen, why, and whether the new process reduced them.
35. What is normalisation in databases, and why should an analyst care?
Normalisation organises a database into separate related tables so each fact is stored once, for example customers in one table and orders in another, linked by a customer ID. It avoids duplication and update errors in applications. Analysts care because they must join these tables back together, and joining on the wrong key or at the wrong grain duplicates rows and inflates totals. Analytical warehouses often use star schemas instead.
Learn more: Normalization, Denormalization and Wide Tables
Case study and experienced data analyst interview questions
36. Case: sales dropped 20% last month. How do you investigate?
First confirm the drop is real: check the data pipeline, definitions and whether last month had fewer days or a holiday. Then break revenue down: number of orders times average order value, and by region, channel, product and customer type, to find where the drop is concentrated. Look for internal changes (price, website, marketing) and external ones (competitors, season). Form a hypothesis, test it with data, and report the cause with a recommended action.
37. Case: how would you measure the success of a new feature?
Start from the feature’s goal, such as faster checkout. Choose one primary metric linked to it (checkout conversion rate), a few supporting metrics (time to checkout) and guardrail metrics that must not get worse (refunds, support tickets). Ideally run an A/B test; if not possible, compare before and after carefully, accounting for seasonality. Agree the metrics and the success threshold before launch.
Also check who used the feature: if only a small group tried it, the overall effect may be too small to see, so measure the effect on users who actually used it as well.
38. How do you decide the sample size for an A/B test?
Sample size depends on four things: the baseline rate of the metric, the smallest effect worth detecting, the significance level (often 5%) and the power (often 80%). Smaller effects and noisier metrics need more users. Use a sample size calculator or a power analysis before starting, then run the test for the planned length, ideally whole weeks, instead of stopping as soon as the result looks significant.
Example: with a 5% conversion rate, detecting a lift to 5.5% needs far more users than detecting a lift to 7%. If traffic is too low, test a bigger change or a more frequent metric.
Learn more: A/B Testing: Design, Sample Size and Analysis
39. What are common mistakes in A/B testing?
Stopping the test early when the result first looks significant (peeking), testing many metrics or variants and reporting only the winner, unequal group sizes caused by a bug (sample ratio mismatch), running too short to cover weekly patterns, and ignoring novelty effects. Also, a statistically significant result can be too small to matter. Plan the analysis in advance and check the groups are balanced.
Also watch for one group seeing a slower page or a bug: always check that the test itself worked before trusting the result.
Learn more: A/B Testing: Design, Sample Size and Analysis
40. What is Simpson’s paradox?
Simpson’s paradox is when a trend appears in each of several groups but reverses when the groups are combined. For example, a treatment can have a higher success rate than another in both mild and severe cases, yet a lower rate overall, because it was given mostly to severe cases. It shows why you should check results by important segments and think about what drives group sizes before drawing conclusions.
41. How do you prioritise many requests from different stakeholders?
Ask what decision each request supports, how valuable it is and how urgent it is, then rank them by impact and effort. Share the list and the reasons with the stakeholders, and agree priorities with your manager when they conflict. Look for requests that can be answered by one reusable dashboard or dataset. Keep people informed when timelines change.
Some requests can be answered in a few minutes with an existing report; doing those quickly while scheduling bigger work keeps stakeholders happy and protects your time for the analyses that matter most.
42. Tell me about an analysis that changed a decision.
Use the STAR structure. Describe the business question, the data you used and how you checked it, what you found (with a number), how you presented it, and what decision changed as a result, with its measured effect if known. For example: “I showed that 60% of returns came from one size chart error; after the fix, returns fell by a third.” Interviewers look for impact, rigour and communication.
Keep the story short and concrete, and be ready to explain how you checked your data and what you would do differently next time.
43. How do you make sure your numbers are correct?
Check data against a trusted source, such as finance totals, before analysing. Check row counts after every join and filter. Look at the data, not just the totals, for odd values. Recalculate key numbers a second way, such as from a different table. Ask someone to review important analyses, document definitions and assumptions, and compare with previous periods to spot unlikely jumps.
A good habit: before sharing, ask yourself “does this number make sense?” If revenue doubled overnight, it is far more likely to be a data error than a real change.
44. What is RFM analysis?
RFM segments customers by recency (how recently they bought), frequency (how often they buy) and monetary value (how much they spend). Each customer gets a score, often 1 to 5, for each dimension, and the combination defines segments such as loyal high-value customers, new customers or customers at risk of leaving. It is simple to build in SQL or Excel and helps target marketing and retention efforts.
Example: a customer who bought last week, buys monthly and spends a lot scores 5-5-5 (a champion); one who has not bought in a year scores low on recency and may get a win-back offer.
45. How do you calculate customer churn and retention?
Define an active customer first, such as someone who bought in the last 90 days, because the definition changes the result. Monthly churn rate = customers who stopped during the month divided by customers at the start of the month; retention rate is one minus churn for the same group. Track it by cohort and by segment, and exclude customers who joined during the month from the starting count.
Example: 1,000 active customers on 1 March, 50 of them stopped by 31 March, so monthly churn is 5% and retention 95%.
Learn more: Product Analytics SQL: Retention, Cohorts, Funnels and Attribution
46. How would you build a sales forecast as an analyst?
Start simple: plot the history and look for trend and seasonality. A baseline can be the same period last year adjusted for growth, or a moving average. Excel’s forecast sheet or a time-series model in Python can add seasonality. Include known future events, such as promotions. Check the method on past data by forecasting periods you already know, report a range rather than one number, and update it as new data arrives.
47. How do you handle a stakeholder who disagrees with your analysis?
Listen to understand their reason; they may know something the data does not show. Walk through your method, data and assumptions openly, and check any specific point they raise. If they are right, update the analysis and say so. If you still disagree, suggest a way to test it, such as a small experiment. Keep the discussion about evidence, not about being right.
Disagreements often come from different definitions, such as what counts as an active customer, so agreeing definitions first prevents many arguments.
48. How do you present an analysis in a five-minute meeting?
Start with the one-sentence answer and the decision you recommend. Show one or two charts that support it, with titles that state the finding (“Returns doubled for size M”) rather than the topic. Mention the main caveat. Keep the appendix ready for questions. Practise it once, and finish by confirming the next step and owner.
Send the slides or a short summary beforehand when possible, so the meeting can be spent on the decision rather than on explaining the numbers. Afterwards, write down what was decided and share it.
49. What is the difference between a metric definition and a dashboard?
A metric definition says exactly how a number is calculated: the source table, filters, time zone, how refunds and cancellations are treated, and who owns it. A dashboard displays metrics. Problems start when two dashboards calculate the same metric differently. Good teams write definitions once, in a shared semantic model or metrics layer, and build all dashboards on them.
Example: “active customer” might mean one purchase in 30 days on one dashboard and 90 days on another, which makes the two dashboards disagree and damages trust in both.
50. What would you do in your first 90 days as a data analyst?
First month: learn the business, the main metrics and the data sources, and meet the key stakeholders. Second month: deliver a few useful analyses quickly, and fix any data quality problems you find. Third month: improve something lasting, such as a trusted dashboard or documented metric definitions, and agree priorities for the next quarter. Throughout, ask questions and write down what you learn.
The aim is to become someone the team trusts for fast, correct answers, and then use that trust to improve how the team works with data.
How to prepare
- SQL is tested in almost every data analyst interview. Practise joins, GROUP BY with HAVING, window functions, CTEs and date handling until you can write them without looking anything up. Start with SQL joins and window functions.
- Excel: pivot tables, XLOOKUP, SUMIFS and COUNTIFS, and basic data cleaning.
- Statistics: averages and spread, distributions, p-values, confidence intervals and A/B testing.
- Practise two case studies out loud: “why did a metric drop?” and “how would you measure a new feature?”.
- Bring one project you can explain end to end: the question, the data, your method, the finding and what changed because of it.