Capital One Assessment Test Applied Techniques 5 — Questions and Answers
Question 1: A dataset shows that income and credit score have a correlation of 0.55. What is the most accurate interpretation?
- Income causes higher credit scores
- There is a moderate positive relationship between income and credit score (Correct answer)
- Income explains 55% of the variation in credit scores
- Higher income always leads to higher credit scores
Correct answer: There is a moderate positive relationship between income and credit score
Correlation of 0.55 indicates a moderate positive association, not causation or a perfect relationship.
Question 2: An analyst needs to identify customers most likely to respond to a balance transfer offer. Which technique is most appropriate?
- Clustering to group customers by demographics
- A regression model predicting response probability (Correct answer)
- A time-series forecast of future balances
- A hypothesis test comparing two group means
Correct answer: A regression model predicting response probability
Logistic regression or similar predictive models estimate the probability of a binary outcome like responding to an offer.
Question 3: A table shows Q1 revenue by product: Checking=$1.4M, Savings=$0.9M, Credit Card=$3.2M, Loans=$2.1M. Credit cards represent approximately what percentage of total revenue?
- 38%
- 42%
- 42.1% (Correct answer)
- 43%
Correct answer: 42.1%
Total = 1.4+0.9+3.2+2.1 = 7.6M; 3.2/7.6 ≈ 42.1%.
Question 4: A risk team uses a scorecard where each attribute contributes points. A customer earns: payment history=35, debt ratio=20, account age=15, credit mix=10. Their total score is 80 out of 100. They are denied if score < 70. What is the outcome?
- Denied — score is below threshold
- Approved — score exceeds the threshold (Correct answer)
- On hold — score needs manual review
- Cannot be determined from given data
Correct answer: Approved — score exceeds the threshold
35+20+15+10=80, which exceeds the denial threshold of 70, so the customer is approved.
Question 5: An analyst discovers that a model trained on 2019–2021 data is underperforming in 2024. The most likely explanation is:
- The model has too many variables
- Concept drift — the relationship between features and outcomes has shifted (Correct answer)
- The 2024 dataset is too large
- The model was evaluated on training data, not test data
Correct answer: Concept drift — the relationship between features and outcomes has shifted
Concept drift occurs when the statistical properties of the target variable change over time, making older models less accurate.
Question 6: A segment analysis shows that 20% of customers generate 78% of revenue. This is an example of:
- Normal distribution
- Pareto principle (Correct answer)
- Regression to the mean
- Simpson's paradox
Correct answer: Pareto principle
The Pareto principle (80/20 rule) describes how a small portion of inputs often drives the majority of outputs.
Question 7: A bank wants to reduce false positives in its fraud detection system without retraining the model. The most direct approach is:
- Collect more training data
- Lower the classification threshold to flag fewer transactions
- Raise the classification threshold to flag fewer transactions (Correct answer)
- Switch to a different model architecture
Correct answer: Raise the classification threshold to flag fewer transactions
Raising the threshold means only higher-confidence predictions are flagged as fraud, reducing false positives.
Capital One Assessment Test Applied Techniques 5
A dataset shows that income and credit score have a correlation of 0.55.
What is the most accurate interpretation?