The short answer
Data quality means data is fit for its intended use: dashboards, ML features, billing, compliance, or ops decisions. "Fit" is not absolute. A table can be perfect for nightly finance and still fail for real-time fraud.
The 6 dimensions interviewers expect:
- 1Accuracy: values match reality / source of truth (order total is correct).
- 2Completeness: required fields and rows are present (no missing
customer_idwhere needed). - 3Consistency: same fact agrees across systems (warehouse revenue matches billing for that day).
- 4Timeliness / freshness: data arrives soon enough for the use case ( age).
- 5Validity: values follow rules and formats (
statusin allowed set; email looks like an email). - 6Uniqueness: no unwanted duplicates at the declared grain (
order_idunique infct_orders).
Example: fct_orders
accuracy: amount matches Stripe charge
completeness: order_id / amount / ordered_at not null
consistency: sum(amount) ≈ Stripe payout for the day
freshness: max(ordered_at) within 1 hour of now
validity: currency in {USD, EUR, INR}
uniqueness: 1 row per order_idInterview tip: Define quality as "fit for purpose," list the 6 dimensions with one concrete check each, and say which dimensions matter most depend on the consumer.