CTE Data Analytics 3 — Questions and Answers
Question 1: Which visualization type is most effective for showing the distribution of call duration across millions of CDRs (Call Detail Records)?
- Histogram with logarithmic y-axis (Correct answer)
- Stacked bar chart by product line
- Pie chart by customer tier
- Scatter plot of revenue vs. call count
Correct answer: Histogram with logarithmic y-axis
A histogram reveals the full shape of the distribution (skew, outliers, modes) across large datasets, which is critical for CDR analysis.
Question 2: A telecom executive wants to reduce customer churn. Which predictive modeling output is most operationally useful?
- A ranked propensity-to-churn score per customer with associated confidence intervals (Correct answer)
- The overall churn rate for last quarter
- A list of all customers sorted alphabetically
- Total revenue lost to churn over 12 months
Correct answer: A ranked propensity-to-churn score per customer with associated confidence intervals
Individual propensity scores allow retention teams to prioritize outreach to the highest-risk customers before they churn.
Question 3: What does a Lorenz curve reveal when applied to telecom revenue data?
- The degree of revenue concentration among the top percentage of customers (Correct answer)
- Average revenue per geographic region
- Network latency distribution by time of day
- Subscriber growth rate over time
Correct answer: The degree of revenue concentration among the top percentage of customers
The Lorenz curve illustrates how much of total revenue is attributable to what share of the customer base, revealing revenue concentration (Gini coefficient).
Question 4: When building a telecom customer lifetime value (CLV) model, which input variable is typically the strongest predictor?
- Historical payment behavior and plan tenure (Correct answer)
- Customer's geographic ZIP code
- Number of calls to customer support
- Device brand preference
Correct answer: Historical payment behavior and plan tenure
Consistent payment history and long tenure directly indicate future revenue reliability and the customer's value horizon.
Question 5: A/B testing in telecom marketing requires which of the following to ensure valid results?
- Random assignment of customers to control and treatment groups with sufficient sample size (Correct answer)
- Testing on the top 100 highest-value accounts only
- Running the test for exactly 48 hours regardless of traffic
- Using historical data as the test group
Correct answer: Random assignment of customers to control and treatment groups with sufficient sample size
Random assignment eliminates selection bias, and adequate sample size ensures the test has sufficient statistical power to detect real differences.
Question 6: In network analytics, what is 'mean time to detect' (MTTD) a measure of?
- The average time elapsed from a fault occurring to it being identified by monitoring systems (Correct answer)
- The average call setup time on a 5G network
- Time to provision a new enterprise circuit
- Average time between scheduled maintenance windows
Correct answer: The average time elapsed from a fault occurring to it being identified by monitoring systems
MTTD quantifies monitoring effectiveness; a lower MTTD means faults are caught faster, reducing the window for customer impact.
Question 7: Which data governance practice is most critical when combining telecom subscriber data with third-party demographic datasets for analytics?
- Ensuring data lineage is documented and consent frameworks are verified before joining datasets (Correct answer)
- Applying the same encryption key to both datasets
- Storing the combined dataset on the same server as billing records
- Using only monthly aggregated data to reduce file size
Correct answer: Ensuring data lineage is documented and consent frameworks are verified before joining datasets
Documented lineage and verified consent protect against regulatory violations (e.g., CCPA, GDPR) when enriching subscriber data with external sources.
CTE Data Analytics 3
Which visualization type is most effective for showing the distribution of call duration across millions of CDRs (Call Detail Records)?