Amazon Data Analytics

Interview Questions
0-3 YOE
22-25 LPA
SQL Questions

1. Write a query to find the second-highest salary from an

employee table.
Solution: To find the second-highest salary, we can use the DISTINCT clause along with the
ORDER BY and LIMIT clauses.

Example:

SELECT MAX(Salary) AS Second_Highest_Salary

FROM Employees

WHERE Salary < (SELECT MAX(Salary) FROM Employees);

• The subquery (SELECT MAX(Salary) FROM Employees) finds the maximum salary.

• The outer query finds the maximum salary that is less than the highest salary.

2. How do you remove duplicates from a table in SQL?

Solution: To remove duplicates, we use the ROW_NUMBER() window function along with a
DELETE statement.

Example:
WITH DuplicateCTE AS (

SELECT *,

ROW_NUMBER() OVER (PARTITION BY EmployeeID, Name, Department ORDER BY

EmployeeID) AS Row_Num

FROM Employees

DELETE FROM DuplicateCTE

WHERE Row_Num > 1;

• The ROW_NUMBER() assigns a unique number to duplicate rows.

• The DELETE operation removes rows where Row_Num is greater than 1.

3. Explain window functions and give an example.

Solution: Window functions perform calculations across a set of table rows related to the
current row. Unlike aggregate functions, they do not collapse rows into a single result.

Example:

SELECT

EmployeeID,

Name,

Salary,

RANK() OVER (ORDER BY Salary DESC) AS Salary_Rank

FROM Employees;

• The RANK() function assigns a rank to each row based on salary in descending order.

• If two employees have the same salary, they will have the same rank, and the next
rank will be skipped.

4. Write a query to find top N customers by total sales.

Solution: To find the top N customers by total sales, we can use the SUM() aggregate
function along with ORDER BY and LIMIT.

Example:

SELECT CustomerID,

SUM(SalesAmount) AS Total_Sales

FROM Sales

GROUP BY CustomerID

ORDER BY Total_Sales DESC

LIMIT 5; -- Replace 5 with the desired N value

• The SUM(SalesAmount) calculates the total sales for each customer.

• The result is ordered in descending order, and the LIMIT clause restricts the result to
the top N customers.

5. What is the difference between WHERE and HAVING?

Criteria WHERE HAVING

Application Level Works on individual rows Works on aggregated results

Aggregate Functions Cannot be used Can be used

Filtering Phase Filters before aggregation Filters after aggregation

Example:

-- Using WHERE

SELECT *

FROM Employees

WHERE Department = 'Sales';

-- Using HAVING

SELECT Department, AVG(Salary) AS Avg_Salary

FROM Employees

GROUP BY Department

HAVING AVG(Salary) > 50000;

• The WHERE clause filters rows before aggregation.

• The HAVING clause filters based on the aggregated result.

If you need further assistance or more questions solved, feel free to ask!

Data Analytics & Business Insights

6. How do you analyze customer churn using data?

Solution: Customer churn analysis identifies the reasons why customers stop doing
business with a company. It helps develop strategies for customer retention.

Steps to Analyze Customer Churn:

1. Data Collection: Gather customer data like demographics, transaction history,

usage frequency, complaints, feedback, etc.

2. Define Churn: Clearly define what constitutes churn, such as no transactions in the
last 6 months.

3. Feature Engineering: Create features like average transaction value, days since the
last transaction, tenure, etc.

4. Exploratory Data Analysis (EDA): Analyze patterns, distributions, correlations, and

potential factors influencing churn.

5. Predictive Modeling: Apply classification techniques like Logistic Regression,

Decision Trees, or Random Forest to predict churn.

6. Customer Segmentation: Segment customers based on demographics, behavior,

and churn risk.

7. KPIs: Monitor churn rate, retention rate, customer lifetime value (CLV), etc.

Example Query (SQL):

SELECT

CustomerID,

COUNT(CASE WHEN TransactionDate >= DATEADD(MONTH, -6, GETDATE()) THEN 1

END) AS Transactions_Last_6_Months

FROM CustomerTransactions

GROUP BY CustomerID

HAVING COUNT(CASE WHEN TransactionDate >= DATEADD(MONTH, -6, GETDATE()) THEN

1 END) = 0;

This query identifies customers who have not made any transactions in the last 6 months.

7. How would you forecast sales for the next quarter?

Solution: Sales forecasting predicts future sales based on historical data. Accurate
forecasts help in decision-making, inventory management, and marketing strategies.

Steps for Forecasting:

1. Data Collection: Gather historical sales data, seasonal trends, marketing

campaigns, and economic factors.

2. Data Preprocessing: Handle missing values, remove outliers, and create time
series data.

3. Exploratory Data Analysis: Analyze trends, seasonality, and cyclic patterns.

4. Model Selection: Use techniques like:

o Time Series Models: ARIMA, SARIMA

o Machine Learning: Random Forest, XGBoost

o Deep Learning: LSTM (if the dataset is extensive)

5. Evaluation: Evaluate model performance using RMSE, MAE, and MAPE.

6. Forecasting: Predict sales for the next quarter and adjust strategies accordingly.

Example (Python with ARIMA):

from statsmodels.tsa.arima.model import ARIMA

# Assuming 'sales_data' is a DataFrame with 'Sales' column

model = ARIMA(sales_data['Sales'], order=(1, 1, 1))

result = model.fit()

forecast = result.forecast(steps=3) # Forecasting for the next quarter (3 months)

print(forecast)

8. Explain A/B testing and how to interpret the results.

Solution: A/B testing compares two versions (A and B) of a variable (like a webpage) to
determine which performs better.

Steps for A/B Testing:

1. Define Objective: Identify the metric to optimize, like click-through rate (CTR) or
conversion rate.

2. Formulate Hypothesis: State the null hypothesis (no difference) and the alternative
hypothesis (there is a difference).

3. Sample Selection: Randomly divide the audience into two groups — Group A
(control) and Group B (test).

4. Experiment Design: Implement the changes in the test group while keeping the
control group unchanged.

5. Statistical Testing: Use t-tests or chi-square tests to evaluate the significance.

6. Analyze Results: If the p-value < 0.05, reject the null hypothesis — the change has
a significant impact.

Example:

SELECT

Version,

COUNT(*) AS Users,

SUM(Conversion) AS Conversions,

(SUM(Conversion) * 1.0 / COUNT(*)) AS Conversion_Rate

FROM AB_Testing_Data

GROUP BY Version;

• The query helps compare the conversion rates for versions A and B.

9. What metrics would you track for Amazon Prime subscription

growth?
Solution: To track the growth of Amazon Prime subscriptions, monitor the following
metrics:

• Subscriber Growth Rate: Percentage increase in subscriptions over time.

• Churn Rate: Percentage of subscribers who cancel their membership.

• Retention Rate: Percentage of users renewing their subscriptions.

• Customer Lifetime Value (CLV): Estimated revenue from a subscriber during their
lifetime.

• Average Revenue Per User (ARPU): Revenue generated per subscriber.

• Engagement Metrics: Frequency of use, purchase behavior, video streaming hours,

etc.

• Conversion Rate: Percentage of free-trial users converting to paid subscriptions.

Example Query (SQL):

SELECT

COUNT(*) AS Total_Subscribers,

SUM(CASE WHEN SubscriptionStatus = 'Cancelled' THEN 1 ELSE 0 END) AS

Cancelled_Subscribers,

(SUM(CASE WHEN SubscriptionStatus = 'Cancelled' THEN 1 ELSE 0 END) * 100.0 /

COUNT(*)) AS Churn_Rate

FROM PrimeSubscriptions;

• This query calculates total subscribers and the churn rate.

10. How do you handle seasonality in sales data?
Solution: Seasonality refers to periodic fluctuations in data due to seasonal factors.
Ignoring seasonality may lead to inaccurate forecasts.

Techniques to Handle Seasonality:

1. Decomposition: Separate time series into trend, seasonality, and residual

components.

2. Seasonal Adjustment: Remove seasonality to analyze the underlying trend.

3. Seasonal ARIMA (SARIMA): Time series model that accounts for seasonality.

4. Moving Averages: Smooth out seasonal variations.

5. One-Hot Encoding: Represent seasons or months as categorical variables for ML

models.

Example (Python for Seasonal Decomposition):

from statsmodels.tsa.seasonal import seasonal_decompose

result = seasonal_decompose(sales_data['Sales'], model='multiplicative', period=12)

result.plot();

• The decomposition shows the observed, trend, seasonal, and residual components.

Data Visualization (Power BI/Tableau)

11. How do you choose the right visualization for different

datasets?
Solution: Choosing the right visualization depends on the purpose of data representation.
Here’s a guide:

• Comparison: To compare categories or groups

o Bar Chart, Column Chart: Categorical comparisons

 o Stacked Bar/Column Chart: Breakdown of sub-categories

• Distribution: To understand the distribution of data

o Histogram: Distribution of a single variable

o Box Plot: Spread and outliers of data

• Trend Analysis: To track data changes over time

o Line Chart: Continuous time-series data

o Area Chart: Emphasizes magnitude over time

• Part-to-Whole Relationship: To analyze proportions

o Pie Chart, Donut Chart: Share of total

o Stacked Bar/Area Chart: Cumulative comparison

• Correlation: To see relationships between variables

o Scatter Plot: Correlation and outliers

o Bubble Chart: Relationship with an additional variable

• Geographical Data: Location-based insights

o Map Visualizations: Regional and spatial data

Example:
In Power BI, use a line chart to analyze monthly sales trends and a bar chart to compare
regional performance.

12. Explain drill-through & drill-down in Power BI.

Solution: Drill-through and drill-down help in analyzing data at different granularity levels.

• Drill-Down: Navigate from a summary level to a detailed view within the same
visual.

o Example: Click on a year in a bar chart to see monthly breakdowns.

o Implementation: Enable drill-down in visuals and use hierarchical fields like

Year ➔ Quarter ➔ Month.

• Drill-Through: Navigate from one page to another with a focus on specific data.
 o Example: Right-click a region in a report to view detailed sales data on a
separate page.

o Implementation: Create a new report page, add a "Drill-through" field, and

set up "Back" navigation.

Example in Power BI:

To analyze regional sales:

1. Create a bar chart by Region.

2. Add a drill-through field like Product Category to view detailed sales by category.

13. What are parameters in Tableau, and how do you use them?
Solution: Parameters in Tableau are dynamic input values used for filtering, calculations,
and control of visualizations.

Use Cases of Parameters:

• Filtering data dynamically (e.g., select top N customers)

• Switching between measures (e.g., Sales vs. Profit)

• Adjusting reference lines

• What-if analysis scenarios

Example:
Creating a parameter to switch between Sales and Profit:

1. Create a Parameter:

o Name: Measure Selector

o Data Type: String

o Values: 'Sales', 'Profit'

2. Create a Calculated Field:

IF [Measure Selector] = 'Sales' THEN [Sales]

ELSE [Profit]

END
 3. Use the calculated field in the visualization, and apply the parameter control.

14. How do you create a year-over-year growth visualization?

Solution: Year-over-year (YoY) growth compares metrics across different years to analyze
growth trends.

In Power BI:

• Use DAX to create measures:

Total Sales = SUM(Sales[Amount])

YoY Growth =

CALCULATE(

[Total Sales],

SAMEPERIODLASTYEAR(Sales[Order Date])

YoY Growth % =

DIVIDE(

[Total Sales] - [YoY Growth],

[YoY Growth],

• Create a line and clustered column chart:

o X-axis: Year

o Columns: Total Sales

o Line values: YoY Growth %

In Tableau:
 • Use Table Calculations:

o Compute sales by year

o Use a quick table calculation: Year over Year Growth

o Adjust compute using “Specific Dimensions” for accurate analysis.

15. How do you optimize a Power BI dashboard for performance?

Solution: Optimizing a Power BI dashboard ensures faster load times and smoother user
experiences.

Best Practices for Optimization:

1. Data Modeling:

o Use a Star Schema instead of a Snowflake Schema.

o Create relationships with integers instead of text.

2. Efficient Measures and Calculations:

o Use SUMX, AVERAGEX judiciously.

o Avoid calculated columns; prefer calculated measures.

3. Reduce Data Load:

o Filter unnecessary data at the source level.

o Use DirectQuery for large datasets if real-time data is needed.

4. Query Optimization:

o Remove unused columns and tables.

o Limit the number of visuals on a single report page.

5. Visual Optimization:

o Disable auto-date/time for large datasets.

o Reduce visuals using complex DAX measures.

o Use aggregations for large fact tables.

Example of Optimized DAX:

Optimized Measure = CALCULATE([Total Sales], REMOVEFILTERS(Calendar[Year]))

• Here, REMOVEFILTERS improves performance by reducing filter context.

Statistics & Probability

16. What is normal distribution, and why is it important?

Solution:
A normal distribution is a symmetric, bell-shaped distribution of data where most of the
observations cluster around the central peak (mean). It is defined by its mean (µ) and
standard deviation (σ).

Key Properties:

• Symmetrical around the mean

• Mean = Median = Mode

• 68% of data lies within ±1σ from the mean

• 95% of data lies within ±2σ from the mean

• 99.7% of data lies within ±3σ from the mean (Empirical Rule)

Importance:

• Many natural phenomena (height, weight, IQ scores) follow a normal distribution.

• It is the foundation for various statistical tests like t-tests and confidence
intervals.

• Assumptions of normality are often necessary for parametric tests.

Example:
If the average IQ score is 100 with a standard deviation of 15:

• 68% of people have IQs between 85 and 115.

 • 95% of people have IQs between 70 and 130.

17. Explain p-value in hypothesis testing.

Solution:
A p-value is the probability of obtaining a result at least as extreme as the observed result,
assuming that the null hypothesis (H₀) is true.

• Low p-value (< 0.05): Strong evidence against H₀ → Reject the null hypothesis.

• High p-value (> 0.05): Weak evidence against H₀ → Fail to reject the null hypothesis.

Interpreting p-value:

• p < 0.01: Very strong evidence against H₀

• 0.01 ≤ p < 0.05: Moderate evidence against H₀

• 0.05 ≤ p < 0.1: Suggestive evidence, but not strong

• p ≥ 0.1: Weak evidence

Example:
In an A/B test, suppose p = 0.03 when comparing two marketing strategies.

• Since p < 0.05, there is significant evidence to reject the null hypothesis that both
strategies perform the same.

18. How do you determine if a dataset is skewed?

Solution:
Skewness measures the asymmetry of the data distribution:

• Right (Positive) Skew: Tail extends to the right; Mean > Median > Mode

• Left (Negative) Skew: Tail extends to the left; Mean < Median < Mode

• Zero Skew: Symmetrical distribution

Methods to Determine Skewness:

• Visualization: Use histograms or box plots to observe skewness.

• Skewness Coefficient:
 o Pearson’s Skewness: Skewness=3(Mean−Median) / Standard Deviation

o If skewness > 0: Right skewed

o If skewness < 0: Left skewed

• Using Python (Pandas):

import pandas as pd

data = pd.Series([5, 7, 9, 10, 15, 22, 30])

print(data.skew()) # Positive value indicates right skew

Example:
A dataset of house prices typically shows right skew because a few expensive houses
inflate the mean.

19. What is the difference between correlation and causation?

Solution:

• Correlation: Measures the strength and direction of the relationship between two
variables. It does not imply causation.

o Ranges from -1 to +1.

o +1: Perfect positive correlation

o 0: No correlation

o -1: Perfect negative correlation

• Causation: Indicates that a change in one variable directly causes a change in

another.

o Requires experimental control to establish causality.

o Not all correlated variables have a causal relationship.

Example:

• Correlation: Ice cream sales and drowning incidents are positively correlated.
However, ice cream consumption does not cause drowning — it's due to summer
temperatures.

• Causation: Increasing the number of study hours leads to better exam scores.
20. Explain the Central Limit Theorem (CLT) in simple terms.
Solution:
The Central Limit Theorem (CLT) states that the distribution of sample means approaches
a normal distribution as the sample size increases, regardless of the original data
distribution.

Key Points:

• Sample size ≥ 30 is generally considered sufficient.

• The mean of the sample means equals the population mean (µ).

• The standard deviation of sample means is called the Standard Error (SE):
SE=σ/n
where σ is the population standard deviation and n is the sample size.

Importance:

• CLT allows us to apply normal distribution techniques to analyze sample means.

• Used in hypothesis testing and confidence interval estimation.

Example:
If we take the average height of random samples of 30 students from a population, the
distribution of those sample means will approximate a normal distribution, regardless of
the population's height distribution.

Python for Data Analytics

21. How do you handle missing values in Pandas?

Solution:
Handling missing values is crucial in data cleaning to ensure data quality. Pandas provides
multiple ways to handle missing data.

Methods:

• Identify Missing Values:

import pandas as pd

df = pd.DataFrame({'Name': ['Alice', 'Bob', None], 'Age': [25, None, 30]})

print(df.isnull()) # Check missing values (True = Missing)

print(df.isnull().sum()) # Count of missing values per column

• Drop Missing Values:

df.dropna(inplace=True) # Drops rows with any missing values

• Fill Missing Values:

df.fillna(0, inplace=True) # Replace missing values with 0

df['Age'].fillna(df['Age'].mean(), inplace=True) # Fill with mean

• Forward/Backward Fill:

df.fillna(method='ffill', inplace=True) # Forward fill

df.fillna(method='bfill', inplace=True) # Backward fill

Choosing the Right Method:

• Drop if missing data is minimal.

• Fill with statistical measures for numerical data.

• Forward/backward fill for time-series data.

22. Write a Python function to calculate the moving average of a

time series.
Solution:
A moving average smoothens time-series data by averaging data points within a specific
window.

import pandas as pd

def moving_average(data, window):

return data.rolling(window=window).mean()
# Example

data = pd.Series([10, 20, 30, 40, 50, 60])

print(moving_average(data, window=3))

Output:

0 NaN

1 NaN

2 20.0

3 30.0

4 40.0

5 50.0

dtype: float64

• The first two values are NaN because the window size is 3.

Use Cases:

• Stock price trend analysis

• Smoothing noisy data in sensor readings

23. Explain groupby() and agg() in Pandas with an example.

Solution:

• groupby(): Groups data based on one or more columns.

• agg(): Applies aggregation functions like sum, mean, max, etc., on grouped data.

Example:

import pandas as pd

data = {'Department': ['HR', 'HR', 'IT', 'IT', 'Finance'],

'Salary': [50000, 55000, 60000, 65000, 70000]}

df = pd.DataFrame(data)

grouped = df.groupby('Department').agg({'Salary': ['mean', 'sum']})

print(grouped)

Output:

Salary

mean sum

Department

Finance 70000 70000

HR 52500 105000

IT 62500 125000

Explanation:

• groupby('Department') groups data by the 'Department' column.

• agg({'Salary': ['mean', 'sum']}) applies mean and sum aggregations.

Use Cases:

• Summarizing sales data by region.

• Aggregating customer transactions by category.

24. How do you merge two datasets in Pandas?

Solution:
Merging in Pandas combines data from different DataFrames based on common columns
or indexes. It is similar to SQL joins.

Merge Methods:

• Inner Join: Default; returns matching rows only.

import pandas as pd

df1 = pd.DataFrame({'ID': [1, 2, 3], 'Name': ['Alice', 'Bob', 'Charlie']})

df2 = pd.DataFrame({'ID': [2, 3, 4], 'Age': [25, 30, 35]})

merged_df = pd.merge(df1, df2, on='ID', how='inner')

print(merged_df)

Output:

ID Name Age

0 2 Bob 25

1 3 Charlie 30

• Left Join: All from left, matching from right.

pd.merge(df1, df2, on='ID', how='left')

• Right Join: All from right, matching from left.

pd.merge(df1, df2, on='ID', how='right')

• Outer Join: All data from both, filling NaN for non-matches.

pd.merge(df1, df2, on='ID', how='outer')

Choosing the Right Merge:

• Use inner for common matches.

• Use left for retaining all left records.

• Use outer when retaining all data is crucial.

25. What is the difference between apply() and map() in Pandas?

Feature apply() map()

Purpose Works on DataFrame or Series Works only on Series

Scope Applies functions element-wise Maps values using a dictionary

Flexibility Can apply complex functions Typically for mapping/replacing

Feature apply() map()

Return Type Series or DataFrame Series

Examples:

• apply() on DataFrame:

import pandas as pd

df = pd.DataFrame({'A': [1, 2, 3], 'B': [4, 5, 6]})

df['Sum'] = df.apply(lambda row: row['A'] + row['B'], axis=1)

print(df)

Output:

A B Sum

01 4 5

12 5 7

23 6 9

• map() on Series:

mapping = {1: 'One', 2: 'Two', 3: 'Three'}

df['Mapped_A'] = df['A'].map(mapping)

print(df)

Output:

A B Sum Mapped_A

01 4 5 One

12 5 7 Two

2 3 6 9 Three

When to Use:

• Use apply() for complex operations.

• Use map() for simple value mappings.