Data quality dimensions and risk

Learning objectives

Connect quality dimensions to business impact and prioritize controls by risk.

Core concepts

Quality is fitness for a declared use, not abstract perfection. Completeness, validity, uniqueness, consistency, accuracy, and freshness answer different questions. A control is justified when the likelihood and impact of failure exceed its operating cost.

rules = {
    "order_id": {"complete": True, "unique": True},
    "amount": {"valid": lambda value: value >= 0},
    "created_at": {"fresh_within_hours": 24},
}
risk_score = likelihood * impact

Guided practice

Create a risk register for a customer-orders dataset. For five fields, name the consumer, failure mode, quality dimension, impact, owner, and preventive or detective control.

Checkpoint

Can a dataset be valid but inaccurate? Why should a quality threshold differ between an executive report and an exploratory notebook?

Reference approach

Tie each rule to a consumer decision and an accountable owner. Prioritize identity, financial, and freshness failures before cosmetic standardization.