Skip to main content

🎯 Mock Interview — Full Simulation

15 Questions Across All Rounds (60 Minutes)​

Instructions: Set a timer. Answer each question aloud or in writing before reading the model answer. This simulates the real interview flow at DecisionTree Analytics.

Round 1: HR Screening (5 minutes)​

Q1 (2 min): "Tell me about yourself."​

Model Answer:

"I'm [Name], a recent graduate from [University] with a focus on data analytics. I've built hands-on skills in SQL, Python, Tableau, and Power BI through academic projects — including a customer churn prediction model and an interactive sales dashboard. What excites me about data analysis is translating raw numbers into actionable business decisions. DecisionTree's end-to-end analytics approach — from data strategy to GenAI — is exactly the environment where I want to start my career."

Scoring: ✅ Under 90 seconds? ✅ Past-Present-Future structure? ✅ Mentioned the company?

Q2 (3 min): "Why DecisionTree Analytics?"​

Model Answer:

"Three specific reasons: (1) Your depth — you go beyond dashboards into ML and GenAI with products like AskNeo. (2) Industry breadth — CPG, Financial Services, E-Commerce means diverse exposure as a fresher. (3) Your culture of 'Innovate. Grow. Belong' — at ~160 employees, I'd learn directly from senior leadership while having structured mentorship."

Round 2: Aptitude (5 minutes)​

Q3 (2 min): A product's price increases by 25%. By what percentage must consumption be reduced to keep expenditure constant?​

Answer:

Reduction = (Increase / (100 + Increase)) × 100
         = (25 / 125) × 100 = 20% ✅

Q4 (3 min): In a group of 100 people, 65 like tea, 45 like coffee, and 30 like both. How many like neither?​

Answer:

At least one = Tea + Coffee - Both = 65 + 45 - 30 = 80
Neither = 100 - 80 = 20 ✅

Round 3: SQL (10 minutes)​

Q5 (3 min): Write a query to find the top 3 highest-paid employees in each department.​

Answer:

WITH ranked AS (
    SELECT name, department, salary,
           ROW_NUMBER() OVER (PARTITION BY department ORDER BY salary DESC) AS rn
    FROM employees
)
SELECT name, department, salary
FROM ranked WHERE rn <= 3;

Q6 (3 min): Find customers who placed orders in January but NOT in February.​

Answer:

SELECT DISTINCT customer_id
FROM orders
WHERE MONTH(order_date) = 1
  AND customer_id NOT IN (
      SELECT customer_id FROM orders WHERE MONTH(order_date) = 2
  );

Q7 (4 min): Calculate month-over-month revenue growth percentage.​

Answer:

WITH monthly AS (
    SELECT DATE_FORMAT(order_date, '%Y-%m') AS month,
           SUM(amount) AS revenue
    FROM orders GROUP BY 1
)
SELECT month, revenue,
       LAG(revenue) OVER (ORDER BY month) AS prev_month,
       ROUND((revenue - LAG(revenue) OVER (ORDER BY month)) * 100.0 /
             LAG(revenue) OVER (ORDER BY month), 2) AS growth_pct
FROM monthly;

Round 4: Python (5 minutes)​

Q8 (2 min): Given a DataFrame df with columns product, region, sales, find the top 5 products by total sales.​

Answer:

top5 = df.groupby('product')['sales'].sum().nlargest(5)
print(top5)

Q9 (3 min): How would you handle a dataset with 15% missing values in the income column?​

Answer:

"First, I'd check if the missing data is random (MCAR) or patterned (MNAR). If random and the column is numerical, I'd use median imputation — df['income'].fillna(df['income'].median()) — because median is robust to outliers. If there's a pattern (e.g., high earners don't report), I'd create a binary indicator column income_missing and use model-based imputation. I wouldn't drop 15% of rows — that's too much data to lose."

Round 5: Tableau (5 minutes)​

Q10 (2 min): "What is a Context Filter?"​

Answer:

"A Context Filter creates a temporary subset of data before other filters are applied. Without it, Tableau applies all filters independently — so 'Top 10 products in Delhi' might first find Top 10 across all cities, then filter to Delhi, giving wrong results. Making the city filter a Context Filter ensures Tableau isolates Delhi data first, then finds the Top 10 within it."

Q11 (3 min): "Explain the difference between FIXED, INCLUDE, and EXCLUDE LOD expressions."​

Answer:

"All three compute at a different granularity than the view. FIXED ignores the view entirely — {FIXED [Region] : SUM(Sales)} always gives regional totals. INCLUDE adds a dimension — increases granularity. EXCLUDE removes a dimension — decreases granularity. For example, if my view is at City level and I use {EXCLUDE [City] : SUM(Sales)}, I get the country total on every city row."

Round 6: Statistics (5 minutes)​

Q12 (2 min): "Explain p-value to a non-technical person."​

Answer:

"Imagine you flip a coin 100 times and get 90 heads. The p-value asks: 'If the coin is fair, what's the probability of getting a result this extreme?' That probability is nearly zero — so we conclude the coin is rigged. In analytics, if p < 0.05, we trust that a pattern is real, not random luck."

Q13 (3 min): You ran an A/B test. Control: 4.0% conversion (n=5000). Treatment: 4.5% conversion (n=5000). p-value = 0.12. What do you tell the product team?​

Answer:

"At p = 0.12, the result is NOT statistically significant at the 95% confidence level. However, the directional improvement of +0.5pp is promising. I'd recommend: (1) extending the test for 2 more weeks to increase sample size, (2) checking if specific segments show stronger effects, and (3) calculating the potential business impact — if each conversion is worth ₹500, the potential gain is ₹12,500/day. I wouldn't roll out yet, but I wouldn't kill it either."

Round 7: ML / Decision Trees (5 minutes)​

Q14 (5 min): "Explain how a Decision Tree works and why Random Forest is better."​

Answer:

"A Decision Tree recursively splits data by asking questions — it picks the feature that reduces impurity the most (measured by Gini or Entropy). For example, 'Is monthly spend > ₹5000?' splits customers into groups with different churn rates. It keeps splitting until leaves are pure or a stopping condition is met.

The problem is overfitting — a deep tree memorizes training data including noise. Random Forest solves this by building hundreds of trees, each on a random subset of data and features. Individual tree errors cancel out through majority voting. The result is higher accuracy and better generalization. The trade-off: Random Forest is less interpretable, but for production predictions, accuracy matters more."

Rounds 8-10: Case Study + Behavioral (20 minutes)​

Q15 (8 min): "A retail client sees 15% revenue decline despite 20% increase in website traffic. Diagnose the problem."​

Answer:

Framework: Revenue = Traffic × Conversion Rate × Average Order Value

Given: Traffic ↑ 20%, Revenue ↓ 15%

Step 1: Calculate implied CR × AOV change
  Revenue = Traffic × CR × AOV
  0.85R = 1.20T × CR_new × AOV_new
  CR_new × AOV_new = 0.85/1.20 = 0.708 of original
  → Combined CR × AOV dropped by ~29%

Step 2: Diagnose - which dropped?
  Check Conversion Rate separately:
  - If CR dropped: Maybe new traffic is low-quality (paid ads 
    with poor targeting, bot traffic, irrelevant keywords)
  - If AOV dropped: Customers buying cheaper items, heavy 
    discounting, product mix shift

Step 3: Likely scenario
  Traffic up 20% from a new campaign → low-intent visitors
  These visitors browse but don't buy (low CR)
  The few who do buy get attracted by discounts (low AOV)

Step 4: Recommendations
  1. Segment traffic: organic vs paid — check CR for each
  2. Audit campaign targeting — are we reaching the right audience?
  3. Check discount strategy — are we cannibalizing revenue?
  4. Create a cohort analysis: new vs returning customer behavior

Scoring Guide​

AreaScore (1-5)
HR answers — structured, under time limit__ / 5
Aptitude — correct and quick__ / 5
SQL — correct syntax, efficient queries__ / 5
Python — clean code, explained approach__ / 5
Tableau — concept clarity__ / 5
Statistics — accurate, business-oriented__ / 5
ML — explained with examples__ / 5
Case Study — structured framework__ / 5
Total__ / 40

35+ → You're ready. Walk in with confidence. 25-34 → Good base. Revise weak areas one more time. Below 25 → Go through the round-specific guides again. Focus on theory first.