Data Quality

Short answer: A company assesses data quality along several dimensions and measures each one separately for the fields that matter. A commonly cited set of six dimensions is accuracy, completeness, consistency, timeliness, validity and uniqueness. Other frameworks use different or longer lists, and there is no single agreed set.

In plain terms, accuracy asks whether a value matches reality, such as the correct delivery address. Completeness asks whether required values are present. Consistency asks whether the same fact agrees across records and systems. Timeliness asks whether data is current enough for its use. Validity asks whether values follow the required format and rules, such as a date in a date field. Uniqueness asks whether each real-world item appears once. Quality depends on the use: a missing phone number matters to the delivery team but not to a sales report. Data quality is the ongoing condition of data and the work of managing it. That work includes preventing errors at the point of entry and monitoring. Data cleansing is the corrective work of fixing records that fall short.

The company identifies the fields that matter most for key processes and reports. For each one, it writes rules, such as every delivery address must include a postal code. Software or simple queries measure the share of records that pass each rule. The data owner sets targets, reviews the results each week and assigns failures to the people who can fix them. The team traces root causes and fixes them at the source. For example, it makes fields mandatory, uses drop-down lists instead of free text or checks entries as people type them.

A company has 10,000 customer delivery addresses, and 600 have no postal code. Completeness of postal codes is therefore 94%. Before automating dispatch, the operations team sets a target of 99%, or 9,900 records. With 9,400 complete, it needs to fill 500 of the 600 gaps. A separate check finds 150 customers recorded twice. That is a uniqueness problem, and the team fixes it by merging the duplicate records. Once the team reaches the target, the dispatch automation prints labels without stopping on missing postal codes.

Poor data quality shows up as wrong invoices, failed deliveries, rework, misleading reports and automations that stop or make errors. Automation and AI act on data at speed, so they spread errors faster than manual processes. Measuring quality on the fields that matter shows leaders where to fix data before launching a system or AI project.

Perfect data is neither achievable nor necessary, so targets should reflect the cost of errors. Measuring quality takes effort. Scores can also mislead if staff fill fields with placeholder values to pass a rule. Improvements fade if the team does not fix the causes at the point of entry. Quality rules need owners who understand the business meaning of the data.