Data Quality: Definition & What It Really Means
Data quality is not about perfection — it's about fitness for purpose. A practical look at what data quality means, how to measure it, and why poor data is one of the most expensive problems in modern organisations.
Definition
Data quality refers to the degree to which data is fit for its intended use — whether that's powering a report, training a machine learning model, or driving an operational process. It is not an absolute standard but a contextual one.
The same dataset can be high quality for one purpose and completely unsuitable for another. A phone number field that is populated but never validated may be acceptable for a marketing newsletter and completely useless for SMS delivery automation. Quality is always relative to use.
This is why data quality frameworks focus on defining dimensions — measurable properties of data that determine its fitness for common use cases — rather than a single universal quality score.
Five dimensions of data quality
Most quality frameworks converge on five core dimensions. Together they form a complete picture of whether data can be trusted for a given application.
Correctness
Data accurately reflects the real-world entity or event it represents. Incorrect values in a product catalogue or customer record cascade into every system that depends on them.
Example
A product listed with the wrong unit of measure creates ordering errors and financial discrepancies.
Completeness
All required fields and records are present. Missing data creates blind spots in analysis, breaks automated workflows, and often indicates upstream capture failures.
Example
A CRM with 30% of customer industry codes missing makes segment-based analytics unreliable.
Timeliness
Data is available and current when it is needed. Stale data is especially dangerous in operational systems where decisions are made on a short time horizon.
Example
Inventory levels updated once per day create stock-out risks in high-velocity warehouse operations.
Uniqueness
Each real-world entity appears exactly once in the dataset. Duplicate records inflate metrics, confuse joins, and create inconsistent downstream views.
Example
Duplicate customer records result in double-counting revenue and sending the same customer two invoices.
Consistency
The same data looks the same across systems and over time. Inconsistent representations of the same value prevent reliable aggregation and comparison.
Example
"United Kingdom," "UK," and "Great Britain" refer to the same country but break country-level reports when not harmonised.
ISO 8000: the international data quality standard
ISO 8000 is the international standard for data quality management. It defines requirements for the exchange, portability, and quality of master data — and is increasingly referenced in supply chain, healthcare, and government contexts where data passes between organisations.
Supply chains
Ensures product master data meets quality standards for cross-organisation exchange.
Healthcare data
Supports interoperability of clinical and patient data across providers.
Data exchange
Defines requirements for portable, trustworthy data assets transferred between systems.
Cross-org collaboration
Sets a shared quality baseline for organisations collaborating on joint data products.
The real cost of bad data
Gartner has estimated the average cost of poor data quality at $12.9 million per year for a typical organisation. The impact shows up in four main areas.
Distorted reporting
KPI dashboards built on poor data report numbers that don't reflect reality. Leadership makes decisions based on a fiction.
ML model degradation
Machine learning models inherit the biases and errors of their training data. Garbage in, garbage out — at scale and at speed.
Eroded trust
Once teams stop trusting data, they revert to gut instinct or build parallel shadow reports. The data investment is effectively wasted.
Compliance risk
Many regulatory frameworks (GDPR, ISO standards, financial reporting) require demonstrable data quality controls.
Common causes
- Manual data entry without validation
- Missing or unenforced data standards
- Fragmented processes across teams and systems
- No ownership of data quality at the domain level
- Legacy system migrations that carry forward historical errors
How Enari can help
- Data quality assessments as part of every engagement
- Data governance frameworks and ownership models
- Pipeline validation and quality monitoring
- Data cleaning and enrichment as part of data engineering
- Analytics readiness reviews before AI/ML projects
Related articles
Ready to start?
Let's build something exceptional
Tell us about your data challenge and we'll find the right solution together — fast, practical, and ROI-focused.