Master Data Quality Dimensions and Metrics Reference
Data quality conversations stall when people use the same words differently. This reference gives a working vocabulary: fourteen dimensions, each with a definition, an example rule written for an ERP master, and a metric formula you can compute and trend.
| Dimension | Definition | Example ERP rule | Metric formula | Typical master |
|---|---|---|---|---|
Completeness | Required attributes hold a real value | Item has unit of measure, item type and product group | Records with all required fields / active records | Item |
Uniqueness | Each real-world entity appears once | No two active vendors share a tax ID | 1 minus (surplus duplicate records / active records) | Vendor |
Validity | Values conform to the allowed format or list | Country is a valid ISO 3166-1 alpha-2 code | Records passing format test / records tested | Customer |
Consistency | Related values agree across fields and systems | Currency matches the country on the vendor | Records passing cross-field test / records tested | Vendor |
Accuracy | Value reflects the real world | Item weight matches a physical weigh check | Correct values in sample / sample size | Item |
Timeliness | Data is current enough for its use | Vendor bank details reviewed in the last 24 months | Records reviewed within window / active records | Vendor |
Integrity | References resolve to existing records | Every BOM component exists in the item master | Valid references / total references | BOM |
Conformity to standard | Naming and coding follow the agreed convention | Part numbers follow the defined segment pattern | Records matching pattern / records tested | Item |
Reasonableness | Values fall in a plausible range | Lead time is between 0 and 365 days | Records in range / records tested | Item |
Usage (activity) | Active records are actually used | Active item has a transaction in last 24 months | Active records with activity / active records | Item |
Lineage and audit | Changes are attributable and dated | Every master change has user and timestamp | Changes with audit entry / total changes | All |
Accessibility | Right users can read and maintain data | Only data stewards can edit vendor bank accounts | Authorised edits / total edits | Vendor |
Routing integrity | Operations reference valid work centers | Every routing step has an active work center | Valid steps / total steps | Routing |
Unit consistency | Units of measure convert across documents | Purchase UoM converts to stocking UoM | Items with valid conversion / items with two UoMs | Item |
Dimension names vary between frameworks (DAMA, ISO 8000 and ISO 25012 use overlapping terms). Use one agreed vocabulary in your organisation. Each metric is a ratio between 0 and 1, or 0 and 100 percent, that you can chart monthly. Choose rules that tie to a business impact, not every possible rule.
Turn the dimensions into a live scorecard
Optional: we can compute these metrics on your ERP data and publish them for your data owners.
Optional. The tool and every export work without this. No spam.
How to use the reference
Search by dimension or by master data domain such as item, vendor, customer, BOM or routing. For each master you govern, pick the three to five dimensions that matter most and write one or two rules for each. Copy the metric formula into your scorecard and compute it monthly. Export the table to Excel to use as the starting point for a data quality rulebook, then add columns for rule owner, tolerance and cleanup action.
The thinking behind the dimensions
Data quality is defined by fitness for use, so each dimension answers a different failure. Completeness asks whether a value exists. Validity asks whether it is in an allowed format. Consistency asks whether related fields agree. Uniqueness asks whether the entity is duplicated. Accuracy asks whether it matches reality, and it is the costliest to measure because it needs a physical or external check. The first four can be tested by queries on the database; accuracy and timeliness need sampling or review workflows.
- Start with completeness, validity and uniqueness, which are cheap to automate
- Sample for accuracy, for example 50 records per month
- Tie each rule to a downstream process it protects
- Publish the metric, owner and target together
Reading and combining metrics
Each metric is a ratio of good records to tested records. A single overall score hides problems, so report by dimension and by master. If you do need a headline number, weight the dimensions by business impact: an invalid unit of measure on an item can stop a receipt, while an unreviewed description rarely does. Set targets per rule, near 100 percent for transaction-blocking rules, and track a trend line so that owners see progress rather than a static grade.
Common mistakes
Teams often define dozens of rules and measure none. Start with ten rules and automate them. Another mistake is measuring on all records instead of active ones, which drowns the signal in obsolete data. Finally, a rule without an owner and a cleanup action is just a report. Assign every rule a data steward and a fix path, whether that is a workflow, a mass change or a training item.
Frequently Asked Questions
How many dimensions do I need?
Most programs start with four to six: completeness, uniqueness, validity, consistency, timeliness and sometimes integrity. Add accuracy sampling later. More dimensions do not help unless you assign rules and owners to them.
What is the difference between validity and accuracy?
Validity means the value is in an allowed format or list, for example a real country code. Accuracy means it is true: the code is the right country for this vendor. A valid value can be inaccurate, so both are needed.
How do I measure accuracy?
Sample a small set of records each month and check them against a physical or authoritative source, such as weighing an item or confirming a vendor address with a registry. Report the share correct and fix errors found.
Which frameworks use these dimensions?
DAMA, ISO 8000 and ISO 25012 describe overlapping sets with slightly different names. The exact label matters less than using one agreed definition across your organisation.
If you want these dimensions turned into scheduled checks against your ERP with owner alerts, we can help set that up on your own infrastructure.
Related Tools
ERP Master Data Completeness Profiling SQL Generator
Enter a table and up to four columns to get SQL that reports null, blank and placeholder rates for each field.
Developer ToolsERP Duplicate Master Record Finder SQL Generator
Fill in your master table and match columns to get a ready-to-run SQL query that groups likely duplicates for review.
Quality & InspectionMaster Data Defect Rate, DPMO and Sigma Calculator
Enter records tested, checks per record and defects found to get defect rate, DPMO, first-pass accuracy and sigma level.
ERP OperationsMaster Data Management Maturity Self-Assessment
Ten questions that place your master data practice on a maturity scale and suggest the next steps.