How is Data Quality Measured?

In the age of information, we continuously have to deal with more and more data. For a contemporary business, data is as necessary as any physical material that is used, if not more.

And as with anything else that businesses use or produce, what matters is quality. However, data quality might seem a little trickier to grasp than the quality of more traditional products and tools. After all, it is not something tangible or visible at first glance. So, how would someone go about evaluating the quality of the datasets they possess?

What value does high-quality data bring to the business?

Data quality, just as the quality of any product or service, can be recognized by how well it does what we expect from it. Therefore, the first step in defining high-quality data is looking at what we want from it.

Generally, data is used, on the one hand, as part of daily operations of a business, and, on the other hand, as part of business intelligence. We use data for business intelligence when we want to make informed decisions about what strategic actions need to be taken to ensure that the company successfully moves forward.

This means that high-quality data will enhance the daily workflow of the business, allowing everything to run smoothly. It will also provide valuable insights into the current situation relevant to the business, which will allow for better management decisions.

Dimensions of data quality

Having this in mind, we can look at the criteria for measuring data quality. A good way to start is the list of data quality dimensions provided by DAMA UK, a not-for-profit data management community.

1. Completeness

When evaluating the quality of a particular dataset, one thing to check for is whether there are important elements missing. For example, if you have requested some information about the customer, but values are missing in the telephone number field, the data is incomplete. However, if the value is missing in a field that you deemed optional, for example middle name, the data is still considered complete as long as you have all the mandatory information in there.

2. Uniqueness

Sometimes also called deduplication, as it concerns whether there are duplicate data units in your database. Duplicates mean lower quality as they take up valuable storage space and give you the wrong impression regarding how much data you actually have.

3. Timeliness

This criterion refers to whether your data is recent enough to still be as useful as expected. The utility of information might decrease gradually with time, or it might drop dramatically after a certain deadline expires. Therefore, the data that is available within the required time frame is of higher quality.

4. Validity

The data is valid when it is the correct type in relation to what was requested. For example, if the field in the dataset is for putting in telephone numbers, the data that is typed in letters instead of digits is invalid. The letters might be meaningful and correct data, for example, the name of the client that was typed instead of the number. But as it is not what was requested, it is not valid for the same purposes.

5. Accuracy

This is the most obvious standard for measuring data quality. Data has to be accurate. No matter how much information is possessed, if it is incorrect, it cannot be considered good information, and it cannot provide much insight for business strategies.

6. Consistency

Information throughout the database should be consistent. This means that there should not be data units that seem to be in conflict with each other. That could happen if, for example, in one dataset within the enterprise a certain telephone number is related to one company name, but on another dataset with a different company name. Even if the information is not inaccurate, which could be if it is the same company that has changed its name, inconsistencies in the way data is recorded might cause confusion and obstruct workflow. Therefore, information that has been updated to avoid such clashes is of higher quality.

There is always more to quality

The set of standards above provides an introduction to how data quality could be measured. However, it is crucial to understand that what constitutes high-quality data may be different in various situations.

This means that, firstly, the list is by no means exhaustive. Other criteria may be added, substitute, or be derived from those above. And, secondly, the definitions of the dimensions are not set in stone. For different purposes, what counts as qualifying for any of the dimensions may be specified or redefined.

At the end of the day, every business needs to figure out what features of data make it most suitable for its purposes. Then the initial set of criteria can be adjusted to reflect what data quality means specifically for this business.