Data Quality Metrics: How to Measure the Accuracy of Your Business Data

Learn the six dimensions of data quality, how to measure each in practice, and how to set thresholds and dashboards that track business data accuracy over time.

You can't improve what you don't measure. That's a principle most business leaders accept readily when it comes to financial performance, sales pipeline, or operational efficiency — but it's far less consistently applied to data quality. Many UK businesses know intuitively that their data is "a bit messy" without having any concrete measure of how messy, which dimensions are worst, or whether things are getting better or worse over time.

Data quality is not a single metric but a multidimensional concept. The widely accepted framework, drawn from ISO 8000 and decades of data management practice, identifies six core dimensions. Understanding and measuring each gives you a meaningful, actionable picture of where your data stands.

The Six Dimensions of Data Quality

1. Accuracy

Accuracy measures whether the data values correctly represent the real-world entities they describe. An email address is accurate if it is the genuine, current email address of the person it's attributed to. A company turnover figure is accurate if it matches the company's most recently filed accounts.

How to measure it: Accuracy is the hardest dimension to measure automatically because it requires comparison against an external source of truth. Approaches include sampling records and manually verifying against Companies House, LinkedIn, or direct contact; comparing your data against a reference dataset from a trusted third-party supplier; or measuring email bounce and hard-fail rates as a proxy for email accuracy.

2. Completeness

Completeness measures the proportion of fields that are populated where they should be. A record with no telephone number or no postcode may be technically valid but commercially useless for certain purposes.

How to measure it: For each field, calculate the percentage of records where that field is populated (non-null, non-blank, non-placeholder). Define which fields are mandatory for your key use cases — for an outbound phone campaign, a phone number is mandatory; for a direct mail campaign, a full postal address is. Report completeness rates per field and per use-case bundle. A threshold of 95% completeness for mandatory fields is a reasonable target for most marketing databases.

3. Consistency

Consistency measures whether the same information is represented in the same way across your dataset, and whether related fields are logically coherent with each other. Inconsistency can exist within a single record (a company record where the postcode doesn't match the stated county) or across records (the same company appearing with three different spellings of its name).

How to measure it: Cross-field consistency checks — does the postcode fall within the stated region? Does the job title align with the seniority band? Does the company registration number match the company name in Companies House? — can be automated. Cross-record consistency requires fuzzy matching to identify similar but not identical representations of the same entity.

4. Timeliness

Timeliness measures whether the data is current enough for its intended purpose. A contact record added five years ago with no subsequent update is likely stale — the person may have changed roles, employers, or contact details. The ICO's guidance under UK GDPR also expects businesses not to hold personal data for longer than necessary, making timeliness a compliance consideration as well as a quality one.

How to measure it: Track the age distribution of your records — what percentage were last updated within the past 12 months, 1–2 years, 2–5 years, and over 5 years? For contact data, research consistently shows that B2B contact information degrades at roughly 25–30% per year as people change jobs, companies restructure, and contacts move on. A record with no update in three years should be treated as potentially unreliable until verified.

5. Validity

Validity measures whether data values conform to the required format, range, or domain for that field. A phone number stored as "please call Monday" is not valid. A date of birth in the future is not valid. A postcode that doesn't match the UK format is not valid.

How to measure it: Define validation rules for each field — regex patterns for phone numbers and postcodes, date range constraints for DOB fields, allowed value lists for categorical fields (e.g. industry sector codes, country codes). Run records through these rules and report the percentage that pass. Validity checking is highly automatable and should be the first quality check applied to any new data import.

6. Uniqueness

Uniqueness measures the degree to which each real-world entity (a person, a company, an address) is represented only once in your dataset. Duplicate records waste marketing budget, frustrate customers who receive multiple copies of the same communication, skew analytical results, and can cause compliance problems if different records for the same individual hold conflicting opt-out statuses.

How to measure it: An exact duplicate rate (records with identical values across all key fields) is easy to calculate. A fuzzy duplicate rate — records that represent the same entity despite minor variations — requires matching algorithms and is more computationally intensive but more meaningful. A well-maintained B2B marketing database should have a duplicate rate below 2–3%; rates above 10% indicate a systemic data entry or integration problem.

Setting Quality Thresholds

Thresholds should be set in the context of your specific use cases and risk tolerance. A financial services firm using data for regulatory reporting needs near-perfect accuracy and validity. An SME running a quarterly email newsletter can tolerate somewhat lower completeness rates without material business impact.

A practical starting point for a UK B2B marketing database might be:

• Email deliverability rate: above 95%
• Postal address validity rate: above 92%
• Mandatory field completeness: above 90%
• Duplicate rate: below 5%
• Records updated within 2 years: above 70%

Dashboarding and Ongoing Monitoring

A one-off quality score is a snapshot; ongoing monitoring is what enables continuous improvement. A data quality dashboard should report each key metric at regular intervals — monthly for most businesses, weekly for those with high-volume ongoing data imports.

Many CRM platforms have basic data quality reporting built in. More sophisticated monitoring can be built using BI tools such as Microsoft Power BI or Tableau, connecting directly to your CRM or data warehouse. The investment in a basic dashboard pays back quickly through earlier detection of data quality degradation.

When to Commission a Data Audit

If you don't have baseline quality metrics, the first step is a formal data audit — a structured assessment that profiles your database across all six dimensions and produces a quality score with actionable findings. A data audit is particularly valuable before a major CRM migration, before launching a significant marketing campaign, after a business acquisition, or when you suspect data quality problems are affecting commercial performance but can't quantify the impact.