Welcome back to Data Quality Foundations! This is lesson three of four, so we are past the halfway mark and building real momentum. In our first lesson, we explored why data quality matters. In our second, we learned to start with the business question and check that our dataset is relevant before touching a single number. Those two skills tell us which data to look at — but they do not tell us how good that data actually is. Today we close that gap by walking through the seven core data quality metrics. By the end, you will have a shared vocabulary for pinpointing exactly where a dataset shines and where it falls short.

It is tempting to label a dataset as simply "good" or "bad," but real data rarely falls neatly into one of those buckets. A dataset might have every row filled in yet contain values that are months out of date. Another might be perfectly current but riddled with typos. Judging quality with a single label hides these important differences.

Instead, we use a set of distinct metrics, each one shining a light on a different aspect of quality. Think of them like the vital signs a doctor checks during a routine visit: temperature, blood pressure, and heart rate each tell a different part of the story, and together they paint a complete picture of health. Our seven data quality metrics — accuracy, completeness, consistency, validity, timeliness, relevance, and reliability — work the same way. Let's examine each one.

While accuracy and completeness focus on individual values, consistency zooms out to check whether the same piece of information agrees across different records, fields, or systems. If a customer's birth date is recorded as 1990-03-15 in the billing system but 03/15/1985 in the support system, at least one of those entries is wrong. Even within a single dataset, inconsistency can surface when the same concept is stored in different formats or units.

Common signs of inconsistency include:

• Mixed date formats (e.g., MM/DD/YYYY vs. YYYY-MM-DD)
• Contradictory values (e.g., a "shipped" order with no shipping date)
• Different spellings for the same item (e.g., "USA" vs. "United States" vs. "US")

Consistency matters because even accurate values become unreliable when they conflict with each other, making it hard to trust any single record.

Timeliness measures whether data is current enough for the decision at hand. A warehouse inventory count from this morning is perfectly timely if we are planning today's shipments. That same count would be dangerously "stale" if it were three months old and we were using it to promise same-day delivery.

One practical way to think about timeliness is the age of the data — the time elapsed between when the data was recorded and when you are actually using it. A short data age is not always required. Historical sales trends, for instance, rely on older data by design. The key question is whether the data is fresh enough relative to how time-sensitive the decision is.

You may recall from our previous lesson that relevance asks whether the data actually pertains to the question we are trying to answer. A dataset can score perfectly on every other metric and still be useless if it measures the wrong thing. We explored relevance in depth last time, so here we will simply note that it remains one of the seven core metrics and deserves a spot in every quality check.

Reliability focuses on whether the data source and collection process are trustworthy and repeatable. If a temperature sensor gives noticeably different readings each time under identical conditions, those readings are unreliable. If a manual data-entry process depends on one person's judgment with no verification step, the resulting data carries more risk. Reliability is typically built over time through consistent processes, documentation, and cross-checks — and it is what separates a one-off lucky dataset from one you can confidently use again and again.

Each metric captures a different dimension of quality, and real datasets often have strengths in some areas and weaknesses in others. The table below summarizes all seven metrics at a glance:

When evaluating a dataset, we rarely look at just one metric in isolation. A dataset might be highly accurate and valid, yet missing so many rows that its low completeness undermines every conclusion we try to draw. Thinking across all seven dimensions gives us a balanced and honest view of where the data stands.

In this lesson, we defined the seven core data quality metrics: accuracy, completeness, consistency, validity, timeliness, relevance, and reliability. Each one highlights a distinct way that data can fall short, and together they form a practical toolkit for evaluating any dataset we encounter.

In our next and final lesson, we will take these metrics further and explore how to decide whether imperfect data is still good enough to support a real business decision. First, though, it is time to put all seven dimensions to work! In the upcoming exercises, you will match each metric to its definition, diagnose real-world data scenarios, and connect common data problems to the business risks they create.