It's important that the data you plan to use is fit for purpose and of the quality and standard that supports good, sound business decisions. Data is only valuable if it's trustworthy and it's only trustworthy if it's of a good quality. To support the assessment of data quality, there is a data quality dimensions tool that helps to determine the validity, accuracy, completely, uniqueness, consistency, and uniformity of the data you're planning to use.

We start with validity. Here we need to ask ourselves if the data we plan to use conforms to the expected format or standard. It needs to align with the defined business rules or constraints. For example, if a student enrols on a course, the age field should contain numerical values. If the field contains non-numerical characters or negative values, the data becomes invalid.

We next look at how accurate the data is and if it reflects true values. In our data set, accurate data means that the age provided for each student is their actual age, the gender is correctly identified, and the grade reflects their performance accurately.

Completeness is where we check whether all required data fields are completed for each record. In our example, each student entry should include their name, age, gender, and grade. If any of these fields are missing for any student, it would indicate the data is incomplete.

We now need to check whether data is duplicated or if a single entity is represented more than once within the data set. In our data set, each student should ideally have a unique name or a unique identifier. If there are duplicate names or identifiers, the data becomes redundant. It's important that data remains the same across different records or databases. For example, if the data set includes a student's age as nineteen in one record and eighteen in another record, without any justifiable reason, it will indicate an inconsistency. Data across different systems must match.

The final step is to check data conforms to a common format throughout the data set. For example, no mixing of numerical and textual representations for the same unit. In our example, all age values should be entered in years. If some ages are listed as months, then the data lacks uniformity, meaning some information could be lost when being analysed. Applying these six dimensions when assessing data quality will ensure you're always working with data that is fit for purpose and that it meets the standards required to support business, data-driven decisions.

Introduction to data quality dimensions