Introduction
Enterprises generate and process enormous volumes of data across applications, departments and business operations. But having access to more data does not automatically mean having reliable data.
An inaccurate customer record can affect revenue reporting. An incomplete asset record can disrupt operations. Inconsistent information across systems can undermine analytics and create problems for regulatory reporting.
This is why enterprise data quality needs more than periodic checks or a dashboard showing an overall quality score.
Data quality dimensions provide a way to assess whether information is fit for a specific purpose. Metrics make those dimensions measurable, while KPIs connect the results to business outcomes. Together, they help organisations identify where data quality is falling short and determine what needs attention.
However, frameworks differ in how they define and group these dimensions. The real challenge is not choosing between six, seven or nine dimensions. It is establishing clear definitions, measurable rules, appropriate thresholds and accountable ownership.
At Edgematics, we approach data quality as an operational capability. By combining data engineering, governance, lineage and automated quality controls, enterprises can move from identifying data issues after they occur to managing quality throughout the data lifecycle.
TL;DR
- Data quality dimensions describe characteristics such as accuracy, completeness and timeliness. Metrics measure whether specific quality rules are being met.
- There is no universally accepted list of dimensions. Enterprises should select those that align with their business requirements and regulatory exposure.
- Quality thresholds should reflect the importance of a data element and the impact of failure, rather than applying a blanket 99% target.
- Composite scores provide an executive overview but can hide individual dimension-level failures.
- Start with critical data elements, automate the checks and establish ownership before expanding the framework.
The Core Data Quality Dimensions
Data quality dimensions are measurable characteristics used to assess whether data is suitable for its intended use.
Most data and analytics teams begin with six commonly used dimensions. Each addresses a different type of data quality issue.
| Dimension | The question it answers | Example quality check |
|---|---|---|
| Accuracy | Does the value reflect the real-world fact? | Customer address matches a verified source |
| Completeness | Is all required information present? | Percentage of customer records with a populated tax ID |
| Consistency | Do different systems agree on the same information? | Subscriber status matches across billing and CRM |
| Timeliness | Is data available when needed and sufficiently current? | Network inventory updated within the agreed window after a fibre rollout change |
| Validity | Does the information follow the required format or domain? | Postal codes and product codes pass defined format rules |
| Uniqueness | Is each entity recorded only once? | Duplicate account rate following an M&A integration |
These six dimensions provide a useful starting point, but they are not the only characteristics used to assess data quality.
For example, the US HealthIT assessment framework includes accessibility, coverage, conformity and integrity alongside several of the dimensions listed above. An academic proposal published in 2024 also suggests expanding the ISO 25012 model with governance, usefulness, quantity and semantics.
The important distinction is that different business environments may require different quality characteristics.
A telecommunications operator may place significant emphasis on the timeliness and uniqueness of network inventory. A financial institution may prioritize accuracy, completeness and validity in customer and transaction records.
The dimensions should therefore reflect how the organisation uses its data and the consequences of getting it wrong.
Why Data Quality Frameworks Differ
There is no single universally accepted definition of data quality or one standard list of dimensions.
Research often groups data quality characteristics into three broad categories:
- Intrinsic: Characteristics of the data itself, such as accuracy and consistency.
- Contextual: Whether the data is suitable for a particular business purpose.
- Representational: Whether the data is presented in a clear and understandable format.
Technology platforms add another layer of complexity. Different tools may use different terminology for similar quality checks. For example, one platform may refer to a check as conformance, while another calls it validity.
This creates a practical challenge for enterprise data teams. If departments use different definitions, quality reports become difficult to compare and governance becomes harder to enforce.
For a Chief Data Officer, the solution is to establish a common data dictionary.
Every dimension should have:
- One agreed definition.
- A clearly identified owner.
- A calculation or measurement method.
- A business context explaining why it matters.
The DAMA Netherlands Dictionary of Dimensions of Data Quality follows a useful approach by framing definitions around the phrase “the degree to which”. This encourages teams to describe each dimension in measurable terms.
Edgematics’ work across enterprise data dictionaries and metadata governance reflects the importance of consistent definitions. When teams work with shared terminology and metadata, they can establish common quality rules across distributed systems rather than managing conflicting standards in isolation.
Data Quality Dimensions vs. Metrics, KPIs and Scores
Dimensions, metrics, KPIs and scores are often used interchangeably. However, each serves a different purpose in an enterprise data quality framework.
Dimension: The characteristic being assessed, such as completeness.
Metric: A numerical or logical measure that determines whether a specific quality rule is being met. Examples include percentages, error counts, true or false checks and error ratios.
KPI: A metric connected to a business outcome. For example, the percentage of billable services with a complete and valid address.
Score: An aggregate measure combining multiple metrics to provide an overall view of data quality.
Consider a telecommunications example.
An operator needs every active fibre service to have a valid premises ID.
- Dimension: Completeness.
- Quality rule: Every active fibre service must have a premises ID.
- Metric: Percentage of active services with a populated premises ID.
- Business KPI: Revenue at risk from services that cannot be billed or provisioned correctly.
The dimension identifies what needs to be assessed. The metric measures whether the rule is being followed. The KPI reveals why the issue matters to the business.
This distinction changes how organisations approach quality management. Instead of measuring data for the sake of reporting, teams can connect quality checks to the business processes that depend on the information.
Our article on why data quality is a revenue problem explores how poor-quality information can affect commercial performance and why organisations need to treat quality as more than a technical concern.
Choosing Data Quality KPIs That Connect to Business Outcomes
A data quality framework should begin with the decisions and processes that depend on the data, not with a predefined list of dimensions.
Three questions help establish the right priorities:
1. Which business processes depend on this data?
Identify the reports, analytical models, operational workflows and regulatory filings that use the information. This establishes where data quality has a direct impact.
2. What does a quality failure cost?
Consider the financial, operational and regulatory consequences. An incomplete customer address may affect billing, while inaccurate asset information may disrupt maintenance or network planning.
3. Who is responsible for resolving the issue?
Every important quality rule needs an accountable owner. Without ownership, dashboards may identify recurring problems without driving corrective action.
The UK Government Data Quality Issues Framework follows a similar approach. It recommends documenting the purpose of each rule, assigning an importance category, assessing the impact of non-conforming data and using the combined assessment to determine an appropriate target range.
This creates a connection between data quality measurement and business priorities.
For enterprises managing complex data environments, Edgematics’ Data Strategy services can help connect data initiatives to business requirements, identify priority use cases and establish the operating approach needed to deliver measurable outcomes.
The goal is to ensure that teams measure the quality characteristics that matter to the business, rather than generating large volumes of technical metrics that do not lead to action.
Setting Data Quality Thresholds That Hold Up
Measuring data quality is only useful when organisations know what level of quality is acceptable and when someone needs to intervene.
A threshold defines the acceptable level of non-conformance for a particular quality rule before corrective action becomes necessary.
A common mistake is applying the same target across every data element. A blanket 99% requirement may look consistent, but it does not account for differences in business importance and risk.
For example, a sanctions-screening field in a bank and a marketing preference field in a retail business do not carry the same consequences when information is incomplete or inaccurate.
The UK Government Data Quality Framework provides a method for determining tolerance by considering the importance of a rule and the impact of failure. For a combined score of 8 to 9, it suggests a target range of 97.5% to under 99.9%.
The HealthIT assessment examples also illustrate that thresholds can differ by dimension, including 85% for accuracy, 75% for conformity and 80% for uniqueness.
These figures should be treated as examples of measurement approaches, not universal enterprise targets.
A practical threshold-setting process should include three controls:
Alert tiers: Define warning and breach levels for each quality rule so teams can distinguish between emerging issues and failures requiring immediate attention.
Escalation ownership: Assign a data steward to investigate warnings and a domain owner or business leader to address significant breaches.
Review cadence: Revisit thresholds when the business context changes. For example, a dataset that begins feeding an AI model may require stricter quality controls than it did when used only for internal reference.
The DAMA Netherlands Data Quality Management System Framework also describes an operating cycle of defining indicators, establishing standards, measuring quality, testing against those standards and taking corrective action.
Edgematics’ approach to data governance and quality management follows the same principle: quality checks need to connect with ownership, monitoring and remediation. This helps enterprises move beyond reporting failures towards managing the conditions that create them.
Data Quality Scoring and Its Limitations
A composite data quality score gives executives a convenient way to understand overall performance. However, relying on one number can hide important problems.
Consider two datasets that both receive an overall quality score of 90%.
The first may have strong accuracy and completeness but poor timeliness. The second may be timely and complete but contain inaccurate records.
Although their overall scores are identical, the underlying risks are very different.
ISO/IEC 25024 defines quantitative measures for the characteristics in the ISO 25012 data quality model. However, it does not prescribe universal ranges for rating levels or grades because these depend on the context of the system and the needs of its users.
For enterprise reporting, we recommend three practices:
- Weight dimensions according to their business impact.
- Document the calculation method and scoring weights.
- Display individual dimension scores alongside the overall score.
This gives leadership a consolidated view without removing the detail needed by data owners and operational teams.
The distinction becomes particularly important when data supports machine learning and AI applications. A dataset with a high overall quality score may still contain a specific weakness that affects model performance.
Our article on building AI-ready data architecture explores why trusted data, governance and traceability need to be considered together when preparing enterprise information for AI use cases.
Building an Enterprise Data Quality Measurement Framework
A list of dimensions and a dashboard of metrics do not constitute a complete enterprise data quality framework.
The organisation needs a consistent operating model that connects quality policies, measurable rules, technical controls, ownership and remediation.
The DAMA Netherlands guidance identifies the importance of a data quality policy that defines which dimensions matter and how the organisation will achieve the required level of quality.
In practice, five components work together.
1. Policy and scope
Identify the critical data elements, the dimensions that apply to each element and the regulatory or business requirements they support.
Not every field requires the same level of control. Prioritising critical data elements helps teams focus on information with the greatest operational, financial or regulatory impact.
2. Quality rules and metrics
Translate each selected dimension into specific quality rules and measurable indicators.
For example, completeness can be measured through the percentage of active customer records with all mandatory fields populated. Accuracy may require validation against a trusted reference source.
Each rule should have a clear purpose, calculation method and acceptable threshold.
3. Controls within data pipelines
Quality checks should operate as data moves through the enterprise environment, not only after information reaches a report or dashboard.
This means embedding validation, deduplication, reconciliation and other relevant checks into ETL/ELT processes.
Edgematics brings data engineering and quality management together by incorporating these controls into the data pipeline. This helps identify issues closer to their source, reducing the effort required for downstream correction and reconciliation.
4. Governance, ownership and lineage
Every important data element and quality rule needs an owner.
Cataloguing and metadata management help teams understand what data exists, while lineage makes it possible to trace information back to its source and follow the transformations it undergoes.
Edgematics’ Data Governance framework explores how decision rights, accountability and automation support governance across enterprise environments.
When these controls work together, a failed quality metric becomes easier to investigate. Teams can identify the source, understand the transformation involved and determine who needs to take corrective action.
5. Reporting and remediation
Quality measurement should lead to action.
Dashboards provide visibility into performance, while issue queues and remediation workflows help teams investigate recurring failures and address root causes.
Reporting should show more than an overall score. It should identify which dimensions are failing, which business processes are affected and whether corrective action is improving the results.
Automation plays an important role here. Edgematics’ work in automating enterprise data operations with AI provides a broader perspective on reducing manual effort and making data operations more repeatable.
Together, these five components turn data quality from a periodic assessment into a continuous management process.
Extending Data Quality Governance to AI
As enterprises introduce machine learning models and AI applications, the consequences of poor data quality extend beyond reporting and operational processes.
Incomplete, inconsistent or poorly governed input data can affect model reliability and the quality of AI-generated outputs.
This makes it important to extend existing data quality controls to the information used by AI systems.
A mature approach should consider:
- The quality and suitability of data used for model training and inference.
- The consistency of definitions across connected sources.
- Traceability into data origins and transformations.
- Ownership and accountability for quality issues.
- Governance controls covering access to data used by AI applications.
The five-pillar model for data governance in AI referenced in the original research provides a useful companion when extending enterprise governance practices to AI environments.
For Edgematics, AI readiness and data quality are closely connected. Building reliable AI applications requires more than deploying models. Enterprises also need governed data pipelines, consistent metadata and quality controls that support the information feeding those applications.
Our Perspective: Data Quality Is an Operational Responsibility
We see a recurring pattern across banks, insurers, retailers and telecommunications operators. Organisations identify the six common data quality dimensions, build dashboards and begin reporting scores. Yet when a quality rule fails, ownership is unclear and corrective action remains limited.
The dimensions themselves are rarely the biggest challenge.
The real work lies in connecting each dimension to a business process, defining a threshold that reflects its importance, assigning accountability and embedding the relevant check where data is produced or transformed.
That is why Edgematics brings data engineering and governance together. Cataloguing, lineage, automated quality rules and pipeline controls need to work as part of the same operating environment.
We also believe that enterprise data quality should start with a focused scope. Select the critical data elements behind a regulatory report, revenue process or priority operational workflow. Establish the measurement and remediation cycle, demonstrate its value and then extend the framework to other domains.
Automation makes this process repeatable. It helps organisations move away from periodic clean-up exercises towards continuous quality management.
How Edgematics Can Help
Building an effective enterprise data quality framework requires more than selecting dimensions and defining metrics. It requires the engineering capabilities and governance processes to apply those standards consistently across the data lifecycle.
Edgematics’ Data Engineering & Governance practice brings together architecture design, ETL/ELT pipelines, data quality management, cataloguing, lineage and compliance frameworks.
We help enterprises translate business requirements into measurable quality rules, embed validation into data pipelines and establish the ownership and traceability needed to manage quality across complex environments.
Whether you are consolidating heterogeneous systems following an acquisition, preparing fibre inventory data for migration, strengthening regulatory reporting or making enterprise data ready for AI, the starting point is the same: identify the data that matters, define what good quality means and make it measurable.
Let’s build an enterprise data quality framework that connects reliable data with better business decisions.
Frequently Asked Questions
What are the six key dimensions of data quality?
The six commonly used dimensions are accuracy, completeness, consistency, timeliness, validity and uniqueness. Other frameworks may include accessibility, coverage, conformity and integrity. Organisations should define the dimensions that best reflect their business requirements.
What are the five Cs of data quality?
There is no single standardized version of the five Cs. Different authors use different lists as memory aids. Each term should be mapped to a clearly defined data quality dimension before it is used for measurement.
How do data quality metrics differ from dimensions?
A dimension describes the characteristic being assessed, such as completeness. A metric measures how well a specific quality rule is met. For example, the percentage of customer records with a populated tax ID is a completeness metric.
Is there a standard threshold for acceptable data quality?
No universal threshold applies to every dataset or dimension. ISO/IEC 25024 leaves rating ranges to the context of each system and its users. Enterprises should set thresholds based on the importance of each rule and the impact of failure.
Why is data quality scoring important?
Scoring helps organisations summarize quality performance across datasets and dimensions. However, an overall score should always be reported alongside individual dimension scores because a composite number can hide specific quality issues.
How can enterprises improve data quality management?
Enterprises can begin by identifying critical data elements, defining measurable quality rules, setting business-aligned thresholds and assigning ownership. Embedding automated checks into data pipelines and establishing lineage, reporting and remediation processes helps make quality management continuous.
