How to Manage Data Quality? A Complete Breakdown of the Six Dimensions

Many companies begin data governance by building standards, metrics, and data catalogs, but when the data reaches the business side the most common question is “Is this data accurate?” Typical symptoms include mismatched sales figures, missing ID digits, duplicate phone numbers, future order dates, or delayed data syncs, which become serious problems as data volume grows.

Effective data‑quality management is not merely cleaning data; it requires a continuous mechanism to discover problems, locate their source, resolve them, and verify the results .

Six Dimensions of Data Quality

1. Completeness – checks whether required data exists. Examples: 5,000 out of 100,000 customer rows lack a region, 3,000 lack a customer type, or an order record is missing the order date. Completeness also covers missing records, such as an ERP system generating 10,000 orders while the downstream warehouse only receives 9,600.

2. Consistency – ensures the same entity has the same representation across systems. A customer may appear as “Shanghai XX Technology Co., Ltd.” in CRM, “Shanghai XX Technology” in ERP, and “XX Technology (Shanghai)” in finance, leading to duplicate aggregation. Consistency problems also arise from differing metric definitions or code mappings, requiring a master data source and unified standards.

3. Accuracy – verifies that data truly reflects business facts. Examples: a customer age recorded as 230 years or an order amount entered as 1.2 million instead of 120 thousand. Accuracy is ensured by (a) comparing with authoritative sources (financial ledgers, master‑data platforms), (b) applying business‑rule validation (price ≥ 0, discount ≤ 100 %), and (c) cross‑data verification (order amount = unit price × quantity).

4. Uniqueness – guarantees a single record per business entity. Duplicate rows such as three identical “Zhejiang ABC Technology Co., Ltd.” entries illustrate primary‑key uniqueness, while composite keys (customer + social‑credit‑code) or entity deduplication handle cases without natural IDs.

5. Timeliness – measures whether data arrives within the required business window. Examples: financial month‑end reports may tolerate T+1, daily operational dashboards need updates before 9 am, e‑commerce monitoring may require minute‑level freshness, and sensor data may need second‑level latency. Rules often specify thresholds like “data must be refreshed within 30 minutes” or “order data must reach the warehouse within 5 minutes of creation.”

6. Validity – checks that data conforms to predefined formats, ranges, and enumerations. Typical checks include phone‑number format, ID‑number length and checksum, email pattern, date parsability, numeric ranges (age 0‑120, discount 0‑100 %), and allowed enum values (order status must be one of ‘Pending’, ‘Paid’, ‘Shipped’, ‘Completed’, ‘Cancelled’).

Five‑Step Data‑Quality Governance Process

Step 1: Identify Critical Data – Prioritize core master data, key business metrics, financial, customer, order, and production data rather than trying to rule every field.

Step 2: Build Quality Rules – Define rules for each dimension (completeness, consistency, accuracy, uniqueness, timeliness, validity) and assign severity levels (e.g., critical for core order amount errors, minor for empty remarks).

Step 3: Automated Monitoring & Anomaly Detection – Deploy automated tasks that flag duplicate order numbers, empty key fields, overdue updates, or cross‑table inconsistencies, and notify owners immediately.

Step 4: Lineage‑Based Root‑Cause Tracing – When an anomaly appears, trace from the abnormal record through processing jobs, upstream tables, and source systems to pinpoint the exact failure point.

Step 5: Close the Loop – Follow the workflow → discover → assign → locate → fix → re‑validate → close, ensuring each issue is resolved, re‑tested, and recorded for future review.

Tools such as FineDataLink 5.0 embody these steps by allowing rule definition at source, full‑linkage monitoring, automatic alerts, lineage analysis, and issue‑tracking dashboards, turning data‑quality from a “problem‑discovery” activity into a closed‑loop governance system.

To evaluate improvement, organizations track metrics like rule‑pass rate, number of detected issues, issue‑closure rate, average resolution time, repeat‑issue rate, and a composite core‑data quality score across the six dimensions.