Data Quality, Testing, and Data Contracts

Question 1

What is data quality in the context of data engineering?

Answer 1

The degree to which data is accurate, complete, timely, consistent, and valid for its intended use.

Question 2

Why must data quality be defined relative to use cases?

Answer 2

Different consumers have different tolerances; data that is acceptable for trend analysis may be insufficient for regulatory reporting.

Question 3

What are the common dimensions of data quality?

Answer 3

Accuracy, completeness, consistency, timeliness, validity, and uniqueness.

Question 4

What does accuracy mean for data quality?

Answer 4

That data values correctly reflect the real-world objects or events they represent.

Question 5

What does completeness mean for data quality?

Answer 5

That required fields and records are present and not missing beyond acceptable thresholds.

Question 6

What does consistency mean for data quality?

Answer 6

That data does not contain contradictory information across systems or within itself and follows agreed rules.

Question 7

What does timeliness mean for data quality?

Answer 7

That data is available within the required time window relative to when events occur or decisions must be made.

Question 8

What does validity mean for data quality?

Answer 8

That data values conform to defined formats, types, ranges, and business rules.

Question 9

What does uniqueness mean for data quality?

Answer 9

That records that should be unique (e.g., primary keys) are not duplicated, avoiding double counting and confusion.

Question 10

What is the role of data quality checks in pipelines?

Answer 10

To automatically detect and surface quality issues early by validating data against expectations as it flows through pipelines.

Question 11

What is a schema check?

Answer 11

A validation that incoming data matches expected column names, types, and structures.

Question 12

What are example content-level data quality checks?

Answer 12

Non-null constraints, range checks, allowed value lists, referential integrity checks, and pattern or format validations.

Question 13

What is anomaly detection at the data level?

Answer 13

Checking for unusual changes in distributions, counts, or patterns that may indicate upstream issues or bugs.

Question 14

Why should data quality checks be automated?

Answer 14

Manual checks don’t scale and are inconsistent, whereas automated checks provide consistent, repeatable validation across runs.

Question 15

What is the difference between hard and soft data quality checks?

Answer 15

Hard checks fail or block the pipeline on violation; soft checks log warnings or raise alerts but allow processing to continue.

Question 16

When might soft checks be preferable?

Answer 16

When some issues are known to occur occasionally and should be monitored but not block the entire pipeline.

Question 17

Why is it risky to use only soft checks?

Answer 17

Critical issues may go unaddressed, leading to bad data in downstream systems and eroding trust.

Question 18

What is a data quality SLA?

Answer 18

An agreement specifying acceptable thresholds for quality metrics such as freshness, completeness, or error rates.

Question 19

Why is monitoring data quality metrics over time important?

Answer 19

It reveals trends, recurring issues, and the impact of changes, making quality improvements measurable.

Question 20

What is a data contract?

Answer 20

A formal, versioned agreement between data producers and consumers specifying schema, semantics, and service levels.

Question 21

What elements typically belong in a data contract?

Answer 21

Schema definitions, field semantics, allowed values, refresh cadence, retention policy, and backward-compatibility rules.

Question 22

How do data contracts reduce breaking changes?

Answer 22

They make expectations explicit and require coordinated schema evolution rather than unannounced changes.

Question 23

What is backward-compatible schema change?

Answer 23

A change that does not break existing consumers, such as adding nullable fields or expanding enums in a controlled way.

Question 24

What are examples of breaking schema changes?

Answer 24

Dropping or renaming fields, changing types incompatibly, or altering meaning without versioning or notice.

Question 25

Why should breaking changes be versioned instead of applied in place?

Answer 25

Versioning allows old and new consumers to coexist, reducing disruption while migrations are completed.

Question 26

What is a contract-first approach to data?

Answer 26

Defining schemas and contracts before implementation and using them as the source of truth for producers and consumers.

Question 27

What is schema registry in the context of data contracts?

Answer 27

A system that stores and validates schema versions for topics or datasets, enforcing compatibility rules.

Question 28

Why is a schema registry useful for streaming systems?

Answer 28

It ensures producers serialize data according to agreed schemas and consumers can deserialize and evolve safely.

Question 29

What is producer-side validation in data contracts?

Answer 29

Checks performed by data producers to ensure they only emit events or records that conform to the contract schema.

Question 30

What is consumer-side validation?

Answer 30

Checks performed by consumers to ensure incoming data matches expectations, guarding against misbehaving producers or drift.

Question 31

What is a data quality firewall?

Answer 31

A layer that blocks or quarantines data that violates critical quality rules before it reaches curated or production tables.

Question 32

What is a quarantine table or error bucket?

Answer 32

A location where bad or suspect records are stored for later analysis and remediation instead of silently dropping them.

Question 33

Why is it important to retain bad records separately?

Answer 33

They provide clues for debugging upstream issues and allow selective correction or replay once problems are fixed.

Question 34

What is referential integrity in data quality?

Answer 34

The guarantee that foreign keys in one table refer to valid primary keys in another, preventing orphan records.

Question 35

Why is referential integrity important for analytics?

Answer 35

Broken references cause missing dimension attributes, inconsistent aggregates, and misleading reports.

Question 36

What is a data quality incident?

Answer 36

An event where data fails to meet agreed quality standards or SLAs, impacting consumers or downstream decisions.

Question 37

Why should data quality incidents be documented and reviewed?

Answer 37

Postmortems identify root causes and process changes that prevent recurrence, improving the system over time.

Question 38

What is the role of ownership in data quality and contracts?

Answer 38

Clear owners for each dataset or contract ensure accountability for quality, evolution, and communication with consumers.

Question 39

How do data quality and contracts relate to trust in a data platform?

Answer 39

Consistent adherence to contracts and visible quality metrics build trust that data is reliable and safe to use.

Question 40

What is a good one-sentence mental model for data quality and contracts?

Answer 40

Define explicit expectations for your data, continuously verify that reality matches those expectations, and make any change to them deliberate, versioned, and visible.