CTO Data Strategy & Analytics 4 — Questions and Answers
Question 1: Which of the following best describes a 'data product' in a modern data strategy?
- A raw database table exported to CSV
- A curated, reliable, discoverable dataset with defined SLAs owned by a domain team (Correct answer)
- A commercial software tool for data visualization
- An ETL pipeline that loads data into a warehouse nightly
Correct answer: A curated, reliable, discoverable dataset with defined SLAs owned by a domain team
A data product is treated like a software product: it has owners, SLAs, documentation, and consumers, ensuring it is trustworthy and reusable.
Question 2: A CTO needs to choose between a relational database and a graph database for a social network recommendation engine. The correct choice is:
- Relational, because SQL is the industry standard
- Graph database, because it natively models entity relationships and traversal queries (Correct answer)
- Document store, because user profiles are JSON objects
- Time-series database, because activity is timestamped
Correct answer: Graph database, because it natively models entity relationships and traversal queries
Graph databases excel at multi-hop relationship traversal, making them ideal for recommendation engines that explore connections between entities.
Question 3: What is the 'data contract' concept gaining traction in modern data engineering?
- A legal agreement between a company and its data vendor
- A formal schema and quality agreement between data producers and consumers (Correct answer)
- An SLA guaranteeing database uptime for analytics workloads
- A GDPR compliance document signed by the DPO
Correct answer: A formal schema and quality agreement between data producers and consumers
Data contracts are machine-readable agreements defining schema, quality expectations, and SLAs between producers and consumers to prevent breaking changes.
Question 4: A CTO is asked to reduce analytics infrastructure costs by 40%. Which approach offers the highest impact without sacrificing capability?
- Deleting all historical data older than one year
- Implementing query cost governance, tiered storage, and workload scheduling (Correct answer)
- Switching all analytics from SQL to Python scripts
- Reducing the number of BI users who can access dashboards
Correct answer: Implementing query cost governance, tiered storage, and workload scheduling
Cost governance through query optimization, auto-scaling, tiered cold storage, and scheduled off-peak workloads typically delivers the largest cloud analytics savings.
Question 5: Which of the following is a leading indicator of poor data governance maturity in an organization?
- Multiple teams report conflicting numbers for the same business metric (Correct answer)
- The data warehouse has more than 500 tables
- Analytics queries take longer than 30 seconds to run
- The company uses more than three different BI tools
Correct answer: Multiple teams report conflicting numbers for the same business metric
Conflicting metrics for the same KPI across teams is a classic symptom of absent or immature data governance, indicating no single source of truth.
Question 6: When implementing data observability, which four pillars are most commonly monitored?
- Latency, throughput, error rate, and saturation
- Freshness, volume, schema, and distribution of data (Correct answer)
- CPU, memory, disk I/O, and network for database servers
- Accuracy, completeness, consistency, and timeliness of ETL jobs only
Correct answer: Freshness, volume, schema, and distribution of data
Data observability frameworks monitor freshness (recency), volume (row counts), schema drift, and statistical distribution to detect anomalies automatically.
Question 7: A company's data science team complains that models trained on historical data perform poorly in production. This is most likely caused by:
- Insufficient computational resources for model training
- Data drift, where the statistical properties of production data have shifted from training data (Correct answer)
- The choice of programming language used for model development
- Excessive feature engineering during experimentation
Correct answer: Data drift, where the statistical properties of production data have shifted from training data
Data drift occurs when the distribution of live data diverges from training data, degrading model performance and requiring retraining or drift detection pipelines.
CTO Data Strategy & Analytics 4
Which of the following best describes a 'data product' in a modern data strategy?