• Accuracy: Does the data reflect reality? (Customer address on file matches their actual address)
• Completeness: Is required data present? (Every product record has a description)
• Consistency: Is the same data represented the same way across systems? ("New Zealand" vs "NZ" vs "nz" in different databases)
• Timeliness: Is the data current enough for the use case? (Yesterday's stock levels: fine for a weekly report; not fine for a live eCommerce site)
• Uniqueness: Are there duplicate records? (Two customer records for the same person with different email addresses)

BA action: When documenting migration or integration requirements, add acceptance criteria for each data quality dimension.

Data governance is the set of policies and processes that control who can access, modify, and use data. BAs encounter data governance when:

• Defining who can see which customer records (access control requirements)
• Documenting retention rules: "customer data must be deleted 7 years after last transaction per Privacy Act 2020"
• Identifying data owners for each domain in a new system (a person who is accountable for data quality in their area)

AI tools (LLMs, recommendation engines, generative systems) produce outputs that require specific types of requirements consideration:

• Accuracy cannot be assumed: AI outputs require human review at defined confidence thresholds. Requirements should specify these thresholds.
• Explainability: In regulated industries, decisions made by AI must be explainable. Define explainability requirements early.
• Bias: AI trained on historical data may perpetuate historical biases. Require bias testing as a non-functional requirement.
• Data provenance: For AI features, document where the training data comes from — this is a Privacy Act concern if it includes customer data.