Research Methodology

Ordinal Scale: What It Is and How to Use It in Research

5 min read

Learn what an ordinal scale is, how it differs from other measurement levels, and the analysis constraints researchers need to know for survey data.

What Is an Ordinal Scale?

An ordinal scale is a level of measurement that categorizes data into groups with a meaningful rank order, but without defined or equal intervals between the categories. It tells you that one value is more, less, or the same as another, without quantifying how much more or less. Education level (high school, bachelor's, master's, doctorate) is ordinal, a doctorate represents more education than a bachelor's, but the "distance" between those degrees isn't a fixed, measurable quantity. Ordinal scales sit between nominal scales (categories without order) and interval scales (ordered categories with equal, measurable gaps) in Stevens' hierarchy of measurement levels.

Why Ordinal Scales Matter in Research

Ordinal data shows up in nearly every survey-based study, from customer satisfaction ratings to employee engagement scores to clinical pain assessments. Mishandling ordinal data, particularly by treating it as interval data without acknowledging the assumption, is one of the most common analytical errors in applied research. A meta-review in Psychological Bulletin found that inappropriate analysis of ordinal data affected the conclusions of roughly 1 in 5 published psychology studies. Knowing what you can and can't do with ordinal data keeps your findings defensible.

How Ordinal Scales Work

What Makes a Scale Ordinal

Three conditions define ordinal measurement:

  1. Mutual exclusivity: Each observation belongs to one category only
  2. Rank order: Categories can be arranged from lowest to highest (or vice versa)
  3. Unknown intervals: The distance between adjacent categories isn't defined or assumed to be equal

That third point is what separates ordinal from interval measurement. On a temperature scale (interval), the difference between 20°F and 30°F is exactly the same as between 70°F and 80°F. On a satisfaction scale (ordinal), the difference between "dissatisfied" and "neutral" isn't necessarily the same as between "satisfied" and "very satisfied."

Common Examples

In market research:

  • Satisfaction ratings (very dissatisfied to very satisfied)
  • Purchase intent (definitely would not buy to definitely would buy)
  • Frequency scales (never, rarely, sometimes, often, always)

In everyday life:

  • Military ranks (private, corporal, sergeant, lieutenant)
  • Restaurant star ratings (1 star to 5 stars)
  • Socioeconomic status (low, middle, high)
  • Race finishing positions (1st, 2nd, 3rd)

The Likert Connection

Likert scales are the most common ordinal scales in survey research. A 5-point Likert item measuring agreement (Strongly Disagree to Strongly Agree) produces ordinal data by definition. The debate over whether Likert data can be treated as interval has been running for decades, with most practitioners landing on a pragmatic middle ground: treat it as approximately interval when you have five or more points, a large sample, and a roughly symmetric distribution.

This pragmatic approach works well enough for business decisions, but if you're publishing academic research or making high-stakes regulatory claims, sticking to ordinal-appropriate methods is the safer path.

Analysis Constraints

The ordinal nature of your data determines which statistics and tests are valid.

Analysis Valid for Ordinal? Notes
Frequency counts Yes Always appropriate
Mode Yes Most common category
Median Yes Middle rank value
Mean Debated Common in practice, technically inappropriate
Standard deviation Debated Assumes equal intervals
Percentiles/quartiles Yes Based on rank position
Spearman's rank correlation Yes Measures monotonic association
Pearson's correlation No Assumes interval data
Mann-Whitney U test Yes Compares two independent groups
Kruskal-Wallis test Yes Compares three or more groups
Chi-square test Yes Tests association between categories
t-test / ANOVA Debated Common in practice, assumes interval

The conservative approach uses only non-parametric methods. The practical approach accepts parametric methods when scales are long enough and samples are large enough. Your choice should match the stakes of the decision you're informing.

Ordinal vs. Other Measurement Levels

Level Order? Equal Intervals? True Zero? Example
Nominal No No No Eye color
Ordinal Yes No No Satisfaction rating
Interval Yes Yes No Temperature (°F)
Ratio Yes Yes Yes Income ($)

Each level up adds a property. Ordinal adds order to nominal. Interval adds equal spacing to ordinal. Ratio adds a meaningful zero to interval.

When to Use an Ordinal Scale

  • Measuring subjective experiences like satisfaction, agreement, preference, or perceived quality where precise quantification isn't possible
  • Creating easy-to-answer survey questions that don't require respondents to assign exact numbers to their feelings
  • Ranking items or options when you need a preference order but not an exact measurement of preference strength
  • Collecting demographic data in grouped formats (income brackets, education levels, age ranges)

Common Mistakes to Avoid

  • Running parametric tests on small ordinal samples without checking whether the approximation holds, with fewer than 30 responses, non-parametric methods are safer
  • Treating the numerical labels as meaningful quantities: the distance between codes 1 and 2 isn't inherently the same as between 4 and 5
  • Using too few categories (2-3 points), which limits variance and reduces the scale's sensitivity to real differences
  • Forgetting to check distribution shape before choosing between parametric and non-parametric analysis

How Quali-Fi Supports Ordinal Scale Analysis

Quali-Fi's platform recognizes ordinal question types automatically and applies appropriate analysis by default. Cross-tabs display median scores and distribution charts rather than only means, and significance testing uses non-parametric methods unless the analyst opts into parametric alternatives. The platform supports Likert scales, ranking questions, and custom ordinal formats across all plans starting at $89/month.

Frequently Asked Questions

Can I calculate a mean for ordinal data?

Technically, no, a mean assumes equal intervals between values, which ordinal data doesn't guarantee. In practice, researchers routinely calculate means for ordinal scales with five or more points, and the results are usually close enough to guide decision-making. Report the median alongside the mean if you want to be thorough.

How many categories should an ordinal scale have?

Between five and seven categories is the sweet spot for most survey applications. Fewer than four limits your ability to detect differences. More than nine overwhelms respondents without adding meaningful precision. The right number depends on how finely your respondents can actually differentiate their experience.

Is a star rating ordinal or interval?

Star ratings (1-5 stars) are ordinal. The difference between 1 star and 2 stars may not represent the same magnitude of quality change as the difference between 4 stars and 5 stars. Platforms that calculate average star ratings are treating ordinal data as interval, a common and generally accepted approximation.

Frequently Asked Questions

Related Guides

Put it into practice

Ready to apply this in your research?

Quali-Fi makes it easy to run surveys, conjoint studies, and more, all in one platform.