What is data quality?

What is data quality?

Data quality is fitness for use. The same dataset may be fine for spotting trends but unsuitable for board reporting.

Here are six commonly used measures for evaluating data quality:

• Accuracy: Values reflect reality. For example, a paid invoice shows the right amount and date.

• Completeness: All required fields are present. For example, there are no missing customer IDs or product SKUs.

• Timeliness: You have data when you need it. For example, the sales team sees daily sales targets before the morning stand-up.

• Consistency: Data aligns across systems. For example, revenue in your data warehouse dashboard matches revenue in your finance report.

• Validity: Values conform to rules and standards. For example, dates parse, currencies use standard formats, and email addresses pass format checks.

• Uniqueness: No duplicated data. For example, one order, one row.

How to protect data quality

Treat your most important fields like a contract. Define the expected type, allowed values, and update cadence. Add tests for grain, time zones, and late backfills. Log rule failures so you can spot drift early.

Know where quality tends to break: manual entry, schema shifts, late-arriving events, joins at the wrong grain, and quiet pipeline failures. Small gaps stack up. A single missing dimension can skew cohort charts, and a stale table can trigger the wrong alert.

Measure data quality directly. Track data freshness per table, null and outlier detection per column, join coverage across keys, and pass rates for validation rules. Review trends weekly, then fix the issues that move business outcomes.

Why data quality matters

Good data quality is the foundation of confident decisions. Without it, leadership meetings turn into debates about whose number is right instead of what to do next.

Here is why data quality matters across the business:

• Trusted decisions: Every team brings the same numbers to the table, so meetings move to action.

• Faster delivery: Clean, up-to-date data makes dashboard and report building reliable and quick.

• Lower risk: Accurate, complete data reduces non-compliance and customer issues.

• Better self-serve: Your team can make confident decisions from consistently defined data, without waiting for someone to pull a number.

Data quality in practice

These examples show where data quality breaks down, what it costs, and how to catch it early.

• Ecommerce orders: A duplicate data load inflates Gross Sales by 3 percent. A uniqueness rule on order_id catches the problem before it reaches executive dashboards.
• Marketing attribution: Incomplete UTM parameters leave gaps in campaign data. A completeness rule flags missing medium and campaign values so the team can correct or drop those rows before budget decisions.
• Revenue reporting: A warehouse refresh runs late. A latency monitor warns the finance team before monthly close so they can pause exports.
• Customer success: A join on account_id changes grain after a schema update. A grain test fails, blocks the deployment, and prevents mismatched data from reaching leadership.

Data quality and metric definitions

Data quality starts at the semantic layer, where your metric definitions live. Define each metric once, mark the trusted version, and tag it so people can find it. Limit editing rights by user, group, and role so definitions don't change without notice.

Apply the same time logic, filters, and groupings everywhere to reduce calculation drift between tools. Surface errors when inputs are null or out of bounds. Set goals and notifications so you know when a metric misses its target or moves outside an expected range.