Certified Data Steward (CDS) — Questions and Answers

Question 1: A data steward is asked to evaluate whether a new AI-driven HR screening tool requires a DPIA under GDPR. Which factor MOST strongly triggers this requirement?

The tool makes automated decisions with significant employment effects (Correct answer)
The tool processes data of more than 1,000 employees
The tool stores data outside the EU
The tool uses cloud infrastructure

Correct answer: The tool makes automated decisions with significant employment effects

GDPR Article 22 and WP29 guidelines identify automated decision-making with significant effects on individuals as one of the clearest triggers for a mandatory DPIA.

Question 2: What does a 'data observability' platform provide beyond traditional data quality monitoring?

End-to-end visibility into data health across pipelines, including freshness, volume, schema, and lineage (Correct answer)
Automatic remediation of all data issues without human intervention
Enforcement of data retention and deletion policies
A replacement for the organization's data catalog

Correct answer: End-to-end visibility into data health across pipelines, including freshness, volume, schema, and lineage

Data observability extends monitoring to provide holistic, continuous insight into data reliability across the full pipeline ecosystem.

Question 3: How should monitoring alerts be configured?

To alert on every minor deviation
With no thresholds to capture everything
Only for critical system failures
With meaningful thresholds that indicate actionable conditions (Correct answer)

Correct answer: With meaningful thresholds that indicate actionable conditions

Meaningful thresholds prevent alert fatigue while ensuring actionable conditions are caught, enabling timely response to genuine issues.

Question 4: In the context of data performance, what does 'cardinality' refer to and why does it matter for indexing?

The number of unique values in a column; higher cardinality makes indexes more selective (Correct answer)
The number of tables in a schema; higher means more joins needed
The depth of a query execution tree; deeper trees indicate poor performance
The number of rows in a table; larger tables need more indexes

Correct answer: The number of unique values in a column; higher cardinality makes indexes more selective

High-cardinality columns have many unique values, making indexes on them highly selective and efficient at narrowing query result sets.

Question 5: Which approach to data classification uses automated scanning tools that inspect content and assign categories based on predefined patterns or keywords?

Policy-only classification
Content-based classification (Correct answer)
Manual classification
User-driven classification

Correct answer: Content-based classification

Content-based classification uses automated tools to scan data content and apply category labels based on pattern matching, keywords, or regular expressions found within the data itself.

Question 6: In a data integration project, 'transformation rules' are BEST documented in which artifact?

Source-to-Target Mapping (STM) document (Correct answer)
Network topology diagram
Server capacity plan
Disaster recovery plan

Correct answer: Source-to-Target Mapping (STM) document

A Source-to-Target Mapping document captures each source field, its target field, and the transformation logic applied during integration.

Question 7: Which data architecture pattern stores integrated, subject-oriented data for enterprise-wide reporting and is the source for downstream data marts?

Data Lake
Operational Data Store (ODS)
OLTP database
Enterprise Data Warehouse (EDW) (Correct answer)

Correct answer: Enterprise Data Warehouse (EDW)

An EDW integrates data from across the enterprise into a single, subject-oriented store that serves as the authoritative source for consistent reporting and feeds departmental data marts.

Question 8: How does poor data quality affect business intelligence?

It improves report formatting.
It results in inaccurate analytics and poor decisions (Correct answer)
It enhances decision-making speed.
It has no impact on analytics.

Correct answer: It results in inaccurate analytics and poor decisions

Poor data quality significantly undermines the effectiveness of business intelligence (BI) initiatives. When data is inaccurate, incomplete, or inconsistent, any analytics performed on it will yield flawed insights and reports. This directly leads to poor, misinformed business decisions, potentially causing financial losses, missed opportunities, and operational inefficiencies.

Question 9: What is the first step in performance optimization?

Add more resources without analysis
Upgrade all hardware immediately
Establish baseline measurements and identify bottlenecks (Correct answer)
Optimize everything simultaneously

Correct answer: Establish baseline measurements and identify bottlenecks

Establishing baselines and identifying bottlenecks ensures optimization efforts target actual performance issues rather than assumed problems.

Question 10: Which metric best indicates that a database system is experiencing I/O bottlenecks?

Increased network packet loss
High memory page faults with fast resolution
High CPU utilization with low wait times
Elevated disk read/write latency and queue depth (Correct answer)

Correct answer: Elevated disk read/write latency and queue depth

High disk latency and long I/O queue depths are the primary signals of an I/O bottleneck limiting query performance.

Question 11: Which Python construct is most efficient for processing large files line-by-line in an automation script without loading the entire file into memory?

file.read()
for line in file: (generator/iterator) (Correct answer)
file.readlines()
pd.read_csv() with low_memory=False

Correct answer: for line in file: (generator/iterator)

Iterating over a file object uses Python's generator protocol, processing one line at a time with minimal memory overhead.

Question 12: A data steward is tasked with establishing data quality thresholds. What should these thresholds be based on?

The maximum thresholds achievable by the source system
Industry benchmarks published by Gartner or Forrester
Business impact analysis and the tolerance level for each use case (Correct answer)
The number of records in the dataset

Correct answer: Business impact analysis and the tolerance level for each use case

Thresholds should reflect how much data quality degradation a specific business process or decision can tolerate before outcomes are materially harmed.

Question 13: What is 'dynamic data masking' and how does it differ from static data masking?

Dynamic masking is used for structured data; static masking is used for unstructured data
Dynamic masking applies masks in real-time based on the requesting user's access rights; static masking permanently transforms data in a copy (Correct answer)
Dynamic masking permanently alters stored data; static masking applies masks on-the-fly during query results
Dynamic masking uses encryption while static masking uses tokenization

Correct answer: Dynamic masking applies masks in real-time based on the requesting user's access rights; static masking permanently transforms data in a copy

Dynamic data masking intercepts query results and applies masking rules based on the requester's privileges in real-time, while static masking creates a permanently masked copy of the dataset.

Question 14: A data steward must ensure that a new marketing analytics platform complies with the CAN-SPAM Act. Which requirement is MOST critical to verify?

All emails must use TLS encryption
Email subject lines must not exceed 50 characters
Every commercial email must include a functioning opt-out mechanism (Correct answer)
Recipients must have opted in before receiving commercial emails

Correct answer: Every commercial email must include a functioning opt-out mechanism

CAN-SPAM requires every commercial email to include a clear, working mechanism that allows recipients to opt out of future messages, and opt-outs must be honored within 10 business days.

Question 15: The 'tone at the top' concept in data governance means:

Senior managers set data retention schedules
Governance policies flow downward from compliance officers
Data quality rules are defined by the most senior technical architect
Executive leadership visibly champions and models data governance behaviors (Correct answer)

Correct answer: Executive leadership visibly champions and models data governance behaviors

Tone at the top means that executive leaders actively champion governance, signal its importance through behavior, and allocate resources — driving organization-wide adoption.

Question 16: What is the relationship between master data and transactional data?

Master data is deleted after each transaction
Master data supports and contextualizes transactional data (Correct answer)
They are the same type of data
Transactional data overrides master data

Correct answer: Master data supports and contextualizes transactional data

Master data and transactional data have a fundamental relationship where master data provides the context for transactional data. Transactional data describes events or activities (e.g., a sale, a payment) and typically references master data entities (e.g., which customer bought which product). Thus, master data supports and contextualizes these transactions, ensuring their accuracy and meaningful interpretation.

Question 17: Which of the following is a key risk when implementing a 'big bang' data migration approach?

Data is migrated incrementally, making rollback difficult
All data is migrated at once, creating a single large window of potential failure (Correct answer)
Source systems continue running in parallel, increasing costs
Migration can only be performed during business hours

Correct answer: All data is migrated at once, creating a single large window of potential failure

A big bang migration moves all data in one cutover event, meaning any errors affect the entire dataset simultaneously with limited ability to recover incrementally.

Question 18: Which type of monitoring checks whether data values fall within expected statistical ranges?

Anomaly or outlier detection monitoring (Correct answer)
Schema monitoring
Referential integrity monitoring
Audit log monitoring

Correct answer: Anomaly or outlier detection monitoring

Anomaly detection monitoring flags values that deviate significantly from historical norms or predefined statistical boundaries.

Question 19: Metadata lineage tracking is most critical for which compliance requirement?

Response time SLAs
Data subject access requests under GDPR (Correct answer)
Software license management
Network throughput auditing

Correct answer: Data subject access requests under GDPR

GDPR data subject access requests require organizations to trace exactly where personal data originated, how it transformed, and where it flows.

Question 20: What does 'data provenance' refer to in data management?

The process of archiving outdated data records
The geographic location where data is stored
The documented history of data's origin, movement, and transformations (Correct answer)
The encryption standard applied to sensitive data

Correct answer: The documented history of data's origin, movement, and transformations

Data provenance tracks where data came from, how it was transformed, and where it has been—critical for auditability and trust.

Question 21: What is a data dictionary's primary function in the context of data classification?

A centralized repository documenting data elements, their definitions, and associated attributes (Correct answer)
A network protocol specification for data transmission
A tool for removing duplicate records from databases
A physical reference book for data backup procedures

Correct answer: A centralized repository documenting data elements, their definitions, and associated attributes

A data dictionary is a centralized repository that documents data elements, their definitions, formats, relationships, and metadata to support consistent data understanding and governance.

Question 22: Which survivorship rule would a data steward apply when multiple source systems disagree on a customer's email address?

Apply a predefined trust score to select the most reliable source's value (Correct answer)
Concatenate all values with a delimiter
Always pick the value from the oldest source system
Leave the field blank until all sources agree

Correct answer: Apply a predefined trust score to select the most reliable source's value

Survivorship rules use source system trust scores, recency, or completeness metrics to elect the best value for the golden record.

Question 23: When a monitoring system sends an alert for a data anomaly, what should the escalation path include?

Automatic deletion of the anomalous records
Public notification to all report consumers about the potential issue
Immediate shutdown of the affected pipeline until investigation completes
Notification to the data steward with context, followed by escalation to the data owner if unresolved within the SLA (Correct answer)

Correct answer: Notification to the data steward with context, followed by escalation to the data owner if unresolved within the SLA

A structured escalation path ensures anomalies are acknowledged and resolved within defined timeframes while keeping appropriate stakeholders informed.

Question 24: A data steward is managing resistance from a senior manager who believes the new data governance program adds bureaucracy. What is the most effective response?

Avoid engaging with the manager and route communications through their subordinates
Escalate to HR for performance management
Reduce governance requirements in the manager's area to avoid conflict
Demonstrate the program's ROI with concrete examples of cost savings or risk reduction relevant to their business area (Correct answer)

Correct answer: Demonstrate the program's ROI with concrete examples of cost savings or risk reduction relevant to their business area

Connecting governance benefits to outcomes the manager cares about — cost, risk, efficiency — converts skeptics by making value tangible and personal.

Question 25: Which approach does passive metadata harvesting use?

Requiring data owners to manually enter descriptions in a catalog
Sampling data at fixed intervals and storing only averages
Waiting for source system administrators to push metadata updates
Automatically scanning systems to extract metadata without interrupting operations (Correct answer)

Correct answer: Automatically scanning systems to extract metadata without interrupting operations

Passive (or automated) metadata harvesting connects to source systems and extracts structural and statistical metadata without manual effort.

Question 26: A data steward is reviewing an issue where transaction records are arriving out of sequence in the data warehouse, causing incorrect balance calculations. This is a failure of which data quality dimension?

Accuracy
Timeliness and consistency (Correct answer)
Accessibility
Uniqueness

Correct answer: Timeliness and consistency

Out-of-sequence delivery violates timeliness (records arrive in wrong temporal order) and consistency (dependent calculations produce inconsistent results).

Question 27: Which data transfer mechanism allows multinational corporations to transfer personal data within their group from the EU without needing separate agreements for each transfer?

Derogations under Article 49
Adequacy Decisions
Binding Corporate Rules (BCRs) (Correct answer)
Standard Contractual Clauses (SCCs)

Correct answer: Binding Corporate Rules (BCRs)

Binding Corporate Rules (BCRs) are approved by a lead supervisory authority and allow intra-group data transfers globally without additional transfer tools.

Question 28: Which of the following best describes 'data provenance' in a governance context?

The business glossary term linked to a data element
The documented history of data origin, movement, and transformation (Correct answer)
The legal jurisdiction governing data storage
The sensitivity classification assigned to a dataset

Correct answer: The documented history of data origin, movement, and transformation

Data provenance tracks where data originated and all transformations it underwent, supporting auditability and trust.

Question 29: A business user reports that a critical KPI dashboard shows NULL values where numbers are expected. After investigating, the data steward finds the issue is in the ETL join logic. Who should be engaged to resolve this?

The ETL developer and the data steward collaboratively (Correct answer)
The data steward alone by rewriting the dashboard query
Only the business user who reported the issue
The database administrator only

Correct answer: The ETL developer and the data steward collaboratively

ETL join logic fixes require collaboration between the ETL developer (technical fix) and the data steward (validating business requirements).

Question 30: When an automated data pipeline completes successfully, which notification mechanism is most appropriate for operational awareness?

Display output on the terminal only
Require a manual sign-off before next run
Write a log entry only
Send a structured alert to a monitoring platform or messaging channel (e.g., Slack, PagerDuty) (Correct answer)

Correct answer: Send a structured alert to a monitoring platform or messaging channel (e.g., Slack, PagerDuty)

Integrating with monitoring platforms ensures the operations team is proactively informed of pipeline status.

Question 31: Which classification category typically applies to an organization's unreleased financial projections and strategic acquisition plans?

Internal Use Only
Public
Unclassified
Confidential (Correct answer)

Correct answer: Confidential

Confidential classification applies to sensitive business information such as financial projections and strategic plans whose disclosure could cause significant harm to the organization or its competitive position.

Question 32: Which challenge is common in master data management?

Data duplication and inconsistency (Correct answer)
Limited user interface options
Excess metadata formats
Too few data sources

Correct answer: Data duplication and inconsistency

One of the most significant challenges in master data management (MDM) is addressing data duplication and inconsistency across various systems and departments. Without a unified approach, the same customer or product information can exist in multiple forms, leading to conflicting data, inaccurate reports, and operational inefficiencies. MDM aims to resolve these issues by creating a single, authoritative view of master data.

Certified Data Steward (CDS)

The Certified Data Steward (CDS) certification validates professional competency in data stewardship across 15 topic areas including data governance, data quality, metadata and master data management, architecture, automation, and regulatory compliance. Offered by eLearningCurve, it is recognized as an industry-standard credential formalizing the Data Steward role.

Questions

35 min

Time Limit

70%

To Pass

Checking for saved progress...

Exam Rules

You can skip questions and return to them later
Flag questions for review before submitting
No feedback shown until you submit the entire exam
Unanswered questions count as wrong — answer everything
10 pretest questions are mixed in and don't affect your score
Timer auto-submits when time runs out
Your progress is auto-saved every 30 seconds