What Is Data accuracy?

Correctness at the field level

A CRM record can combine a correct company domain with an outdated job title and an incorrect phone number. Calling the whole row accurate hides those differences. Define which fact each field represents and compare it with appropriate evidence. The relevant question is whether the value reflects reality for the decision being made, not merely whether the database contains a nonempty string.

Accuracy is separate from consistent formatting, completeness and recency. A recently imported value can be wrong, and a historically correct value can become stale. A standardized title may also be a derived category rather than the person's actual published title. Preserve those distinctions so a convenient label does not become an unsupported claim about the contact or account.

Evaluate a source on a defined sample

Before relying on a provider, choose representative records from the intended market and document how each important field will be checked. Include difficult cases rather than only obvious matches. Keep incorrect, unsupported, ambiguous and stale outcomes visible. An overall percentage without a sampling method, field definition or evidence date is difficult to interpret and even harder to compare with another vendor's claim.

Make corrections traceable

When an employee changes companies, avoid copying the new employer into every historical record without reviewing the relationship and ownership. Preserve the prior value and source where appropriate, record the correction and identify dependent workflows. A change to one contact field can affect account matching, territory assignment and active campaign eligibility in different ways.

Use accuracy evidence alongside provenance, currentness and protected-status checks. A correct address is not automatically an eligible prospect, and a high-quality sample does not certify every row returned later. The practical aim is explainable field-level decisions with a way to challenge and correct errors, rather than a single database badge that later users cannot inspect.