The short answer. Measure data quality across a set of dimensions, completeness, accuracy, consistency, validity, uniqueness, and timeliness, give each a concrete metric, set a threshold for what "good enough" means, and track the results on a scorecard. The scorecard is what turns quality from an opinion into something leaders can manage.

Most teams know their data has problems. Far fewer can say how big those problems are, whether they are getting better or worse, or which ones to fix first. Measurement is what closes that gap. Here is how to do it without drowning in metrics.

Start with the dimensions

Data quality is not one number. It is several distinct qualities, and each one is measured differently. The widely used dimensions are:

Completeness: are the values that should be present actually there?
Accuracy: do the values reflect the real world correctly?
Consistency: does the same fact agree across systems and records?
Validity: do values follow the required format and rules?
Uniqueness: is each real-world thing represented once, without duplicates?
Timeliness: is the data current enough to be useful?

Turn each dimension into a metric

A dimension becomes measurable when you attach a concrete calculation to it. Some practical examples:

Dimension	Example metric
Completeness	Percentage of records with all mandatory fields populated
Accuracy	Percentage of records matching a trusted reference or passing a business rule
Consistency	Percentage of records where a value agrees across systems
Validity	Percentage of values conforming to the required format or range
Uniqueness	Percentage of records with no duplicate
Timeliness	Percentage of records updated within the expected window

Set thresholds that mean something

A metric with no target is just a number. Thresholds define what acceptable looks like and trigger action when data falls below it. Set them by business impact, not by ambition: a field that blocks a migration or breaks a payment deserves a high bar; a nice-to-have attribute does not. Agree the thresholds with the people who own the process, so the numbers carry weight.

Build a scorecard leaders will use

The final step is presentation. A scorecard rolls the metrics up into a view that a non-specialist can read at a glance: current scores by dimension, whether each is above or below threshold, and the trend over time. The trend matters as much as the score, because it shows whether the situation is improving or drifting. A good scorecard answers three questions instantly: how good is our data, where is it worst, and is it getting better?

Leading and lagging indicators

The most useful scorecards mix two kinds of metric. Lagging indicators tell you the current state: how many records are complete, how many duplicates exist right now. Leading indicators hint at where quality is heading: the rate at which new duplicates are being created, or how often validation catches errors at entry. Lagging metrics tell you where you are; leading metrics tell you where you are going. A programme that watches only lagging metrics is always reacting; one that also watches leading metrics can get ahead of problems.

Common measurement mistakes to avoid

Measuring everything. Hundreds of metrics is not rigour, it is noise. Pick the dimensions and fields that carry real business impact and measure those well.
Scores with no owner. A metric nobody is accountable for does not improve. Each key measure needs a person or team who owns the number.
Snapshots without trend. A single reading cannot tell you whether things are getting better or worse. Track over time.
Technical checks only. A field can be correctly formatted and still be wrong. Business-rule checks are what catch data that is valid but not accurate.

Turning many metrics into one clear signal

A common failure of data quality measurement is producing so many numbers that no one can see the wood for the trees. The answer is a deliberate hierarchy. At the bottom sit the detailed metrics, one per rule per field. These roll up into a score per dimension, completeness, accuracy, and so on, for each data domain. Those in turn roll up into a single headline score per domain that a leader can read in seconds. The detail is still there for the people who need to act on it, but the top of the hierarchy answers the only question an executive usually has: is our data getting better or worse?

The art is choosing how to weight the roll-up. Not every rule matters equally, so a simple average can hide serious problems behind a comfortable number. Weighting by business impact, so that a failure on a migration-critical field counts for more than a failure on a cosmetic one, keeps the headline score honest.

Measurement only matters if it drives action

A scorecard that no one acts on is decoration. The measurement has to be wired to a response: a threshold breach should trigger a defined action, whether that is a cleanse, an investigation, or a conversation with the team creating the data. The most effective programmes close this loop tightly, so that a red number on the scorecard reliably leads to someone doing something about it. When that loop is missing, quality measurement becomes a reporting ritual that documents decline without arresting it.

Start with a baseline, then track the trend

The first measurement you take is the most important, because it is the baseline everything else is judged against. Before any improvement work begins, profile your key data and record where each dimension stands. That baseline does two things: it tells you which problems are worst and therefore deserve attention first, and it gives you the reference point that lets you prove, later, that the work made a difference.

Without a baseline, improvement is invisible, and invisible improvement is hard to fund. With one, every subsequent measurement becomes a story of progress or drift that leaders can actually follow, which is what keeps data quality on the agenda rather than quietly falling off it.

Make the numbers a shared language

The final benefit of measuring data quality well is that it gives the business and IT a shared language. Without numbers, conversations about data quality tend to be anecdotal and circular: one side insists the data is fine, the other points to a bad experience, and nothing gets resolved. A scorecard replaces opinion with evidence. It lets a business owner point to a specific completeness score on a specific field and ask why it is below target, and it lets IT show, month over month, that an improvement effort is working. That shared reference is often worth as much as the measurement itself, because it turns data quality from a source of friction between teams into a problem they can look at together and solve.

How deKorvai helps

deKorvai profiles data automatically to discover its structure and content, applies configurable business and technical rules to test each dimension, and presents the results as real-time scorecards with trend insight. Instead of a one-off audit, quality becomes something you can watch continuously and act on as it changes.

Key takeaways

Quality is multi-dimensional: measure completeness, accuracy, consistency, validity, uniqueness, timeliness.
Metrics need thresholds: a number with no target drives no action.
Set the bar by business impact, not by ambition.
Trend beats snapshot: a scorecard should show direction, not just a score.

Frequently asked questions

What are the main data quality dimensions?

The commonly used ones are completeness, accuracy, consistency, validity, uniqueness, and timeliness. Each is measured with its own metric because they describe different qualities of the same data.

What is a data quality scorecard?

A scorecard rolls quality metrics into a single view showing current scores by dimension, whether each is above or below its threshold, and the trend over time, so leaders can see the state of the data at a glance.

How do I set data quality thresholds?

Set them by business impact. Fields that block critical processes such as a migration or a payment deserve a high bar, while lower-impact attributes can tolerate more. Agree the thresholds with the process owners.

How often should data quality be measured?

Continuously where possible. A one-off audit tells you the state on one day; ongoing measurement shows whether quality is improving or drifting, which is what actually drives action.

How to Measure Data Quality: Metrics, Scorecards and Thresholds