What are the key dimensions of data quality, and how do you measure each one? Can you give examples of how poor data quality in each dimension impacts business decisions?

Question

Accepted Answer

Data quality is multi-dimensional, and understanding each dimension enables targeted monitoring and improvement. Poor quality in any single dimension can undermine business decisions, but the specific risks differ by dimension.

Accuracy measures whether data values correctly represent the real-world entity they describe. A customer's address, a product's price, or a transaction's amount must reflect reality. I measure accuracy by comparing data against authoritative sources through periodic audits and automated validation rules. For example, comparing stored postal codes against a postal code reference database. Inaccurate data leads to failed deliveries, incorrect pricing, and financial reporting errors. In BI, inaccurate revenue data flowing into executive dashboards can lead to strategic decisions based on a false picture of business performance.

Completeness measures the proportion of required data values that are present versus missing. I calculate completeness rates at the column level and set thresholds based on business criticality. An email address field with seventy percent completeness means thirty percent of customers cannot be reached by email campaigns, directly reducing marketing effectiveness. I distinguish between data that is genuinely missing versus data that is intentionally null, as these have different implications.

Consistency measures whether the same data represented in multiple places agrees. A customer's name in the CRM should match their name in the billing system. Revenue reported by the finance team should match revenue reported by the sales team if they use the same definition. Inconsistency is the most visible data quality problem because it surfaces when different stakeholders compare notes and find conflicting numbers. I measure consistency through cross-system reconciliation checks that flag discrepancies.

Timeliness measures whether data is available when needed and whether it reflects the current state of reality. A dashboard refreshed daily is timely for strategic reporting but not for operational monitoring. I measure timeliness by tracking the lag between when an event occurs and when it appears in the analytics system. Stale data in a BI dashboard can lead to decisions based on outdated information. If a pricing change took effect yesterday but the dashboard still shows old prices, a sales manager might quote incorrect prices.

Validity measures whether data values conform to their defined format, range, and business rules. A date field should contain valid dates. A percentage should be between zero and one hundred. A status field should only contain allowed values. I implement validity checks as automated constraints and validation rules in the data pipeline. Invalid data often indicates a bug in the upstream system or a data entry error.

Uniqueness measures the absence of unwanted duplicates. Duplicate customer records inflate customer counts, skew per-customer metrics, and cause confusion when multiple records exist for the same entity. I measure uniqueness by running deduplication algorithms on key entities and tracking the duplicate rate over time. Master data management processes address uniqueness at scale.

I prioritize quality improvement efforts based on business impact. A five percent accuracy issue in a high-volume transactional dataset might cost millions, while a twenty percent completeness issue in a low-priority field might have minimal impact. The quality management program should focus resources where the business impact of improvement is greatest.