Nominal, ordinal, interval, and ratio are the four levels of measurement used in research and surveys. Nominal scales categorize data with no order (e.g. gender, product type). Ordinal scales rank data but without equal spacing between points (e.g. satisfaction ratings). Interval scales have equal spacing but no true zero (e.g. temperature in Celsius). Ratio scales have equal spacing and a true zero, enabling the full range of mathematical operations (e.g. age, income).
What Are Levels of Measurement and Why Do They Matter for Surveys?
Every question in a survey collects a particular type of data. That data type — its level of measurement — determines which statistics you can legitimately run, which charts make sense, and ultimately how meaningful your conclusions are.
The four levels were first formalized by psychologist Stanley Smith Stevens in 1946. They form a hierarchy: each level retains the properties of the one before it and adds something new. Getting the level right matters for two reasons:
- Analysis accuracy. Calculating an average on data that only has rank order, for example, produces a misleading number.
- Question design. Knowing which level you need helps you write response options that will actually support your research goal.
The four levels are: nominal, ordinal, interval, and ratio.
Nominal Scale
Definition
A nominal scale assigns data to named categories with no inherent order. The values are labels. You cannot say one category is higher, lower, better, or worse than another — they are simply different.
Key characteristics
- Categories are mutually exclusive
- No ranking or ordering between categories
- No arithmetic operations (you cannot add or average nominal values)
- The only meaningful statistic is the mode (the most common category)
Survey examples
- "What is your gender?" (options: male / female / non-binary / prefer not to say)
- "Which product category did you purchase?" (electronics / clothing / home goods / food)
- "What is your country of residence?"
- "How did you hear about us?" (social media / email / word of mouth / search engine)
In a survey, nominal data is typically collected with single-select or multi-select checkboxes. When you analyze results, you report counts and percentages by category — not averages.
Ordinal Scale
Definition
An ordinal scale places data into categories that have a meaningful order or ranking, but the gaps between the categories are not necessarily equal. You know that one response ranks higher than another, but you cannot say by exactly how much.
Key characteristics
- Categories have a clear rank or order
- Distances between ranks are not assumed to be equal
- You can report the median and mode, but not the mean
- Comparisons like "more than" or "less than" are valid; arithmetic is not
Survey examples
- "How satisfied are you with our service?" (very dissatisfied / dissatisfied / neutral / satisfied / very satisfied)
- "How likely are you to recommend us?" (Net Promoter Score 0–10 scale)
- "Rate the importance of the following features." (not important / somewhat important / very important)
- "What is your highest level of education?" (high school / bachelor's degree / master's degree / doctorate)
The space between "neutral" and "satisfied" may not be the same emotional distance as between "satisfied" and "very satisfied." That unequal spacing is what separates ordinal from interval data.
Interval Scale
Definition
An interval scale has all the properties of an ordinal scale, plus the gaps between values are equal and meaningful. However, an interval scale does not have a true zero point. Zero is either arbitrary or does not exist as a concept on the scale.
Key characteristics
- Categories have order
- Spacing between points is equal
- No true zero — zero does not mean "none of the thing"
- You can calculate the mean, median, and mode
- Ratios are not meaningful (you cannot say "30°C is twice as hot as 15°C")
Survey examples
- Likert-type rating scales — a five- or seven-point scale from "strongly disagree" to "strongly agree" is often treated as interval in practice, though technically it is ordinal (see note below)
- Temperature (Celsius or Fahrenheit) — the classic textbook example; 0°C does not mean "no temperature"
- Calendar years — the year 0 does not mean "no time"
- IQ scores — the zero point is not absolute
A note on Likert scales and surveys: In strict measurement theory, Likert scales are ordinal. In practice, many researchers treat five- and seven-point scales as interval data because the intervals are assumed to be approximately equal. This is a pragmatic choice — one you should make consciously and note in your methodology.
Ratio Scale
Definition
A ratio scale has all the properties of an interval scale, but it also has a true, meaningful zero. Zero means the complete absence of the thing being measured. This makes it possible to make ratio comparisons — you can say one value is twice or half another.
Key characteristics
- Categories have order
- Spacing between points is equal
- True zero exists and is meaningful
- All arithmetic operations (addition, subtraction, multiplication, division) are valid
- You can calculate the mean, median, mode, standard deviation, and more
- Ratios are meaningful: "40 purchases is twice as many as 20 purchases"
Survey examples
- "How old are you?" (age in years — 0 means newborn)
- "What is your annual household income?" (0 means no income)
- "How many times have you purchased from us in the last 12 months?"
- "How many employees does your company have?"
- "How many hours per week do you spend on this task?"
Ratio data gives you the most analytical flexibility. When you collect it in a survey, you can run the widest range of statistical tests.
Comparison: Nominal vs Ordinal vs Interval vs Ratio
| Property | Nominal | Ordinal | Interval | Ratio |
|---|---|---|---|---|
| Has order? | No | Yes | Yes | Yes |
| Equal intervals? | No | No | Yes | Yes |
| True zero? | No | No | No | Yes |
| Statistical operations allowed | Mode, frequency counts | Median, mode, percentile rank | Mean, median, mode, standard deviation | All operations including ratios |
| Survey example | Product category, country | Satisfaction rating, NPS | Likert-type rating scales | Age, income, number of purchases |
How to Choose the Right Scale for Your Survey Question
Picking the right scale is not an abstract exercise — it directly affects whether you can answer your research question. Work through these steps:
1. Define what you want to measure.
Are you categorizing respondents (nominal), ranking their preferences (ordinal), measuring degree of agreement (interval), or counting something concrete (ratio)?
2. Think about the analysis you need.
If you need to calculate averages, you need at least interval data. If you need to say "Group A purchased three times more than Group B," you need ratio data. If you only need to know which category is most common, nominal is sufficient.
3. Match the question format to the level.
- Checkboxes and dropdown lists → typically nominal
- Star ratings, ranking questions, Likert scales → ordinal (or treated as interval)
- Numeric open-text fields → ratio (or interval, depending on what the number represents)
4. Avoid forcing a lower level when you could collect a higher one.
For example, instead of asking respondents to select an age range (ordinal), ask for their exact age (ratio). You can always group ratio data into ranges later — but you cannot recover the original precision once you have collected ordinal buckets.
5. Be consistent with your response labels.
If you use an ordinal scale, make sure the labels form a clear, unambiguous order. Asymmetric or ambiguous labels — like mixing "very satisfied" with "somewhat agree" — distort your data.
How onlinesurvey.ai Helps You Get This Right
Choosing between measurement levels is one of the most common sources of survey design errors — and one of the hardest to fix after data collection has started.
When you describe your research goal in plain language on onlinesurvey.ai, the AI builds your survey questions and automatically selects appropriate question types and response scales. Rather than deciding between a Likert scale, a numeric input, or a category list by hand, you describe what you want to learn — and the platform proposes question formats matched to that intent.
Once responses are collected, the AI generates an executive summary with key findings, patterns, and confidence levels — so the analysis reflects the actual measurement level of your data, not a misapplied shortcut.
FAQ
What is the difference between nominal and ordinal scales?
A nominal scale groups data into unordered categories — such as colors, countries, or product types — where no category ranks above another. An ordinal scale also groups data into categories, but those categories have a meaningful order or ranking, such as "low, medium, high." The key difference is that ordinal data tells you direction, while nominal data only tells you difference.
Can you calculate a mean from ordinal data?
Strictly speaking, no. Ordinal data tells you rank order but not the equal distance between ranks — so averaging the numbers attached to those ranks is mathematically questionable. In practice, many researchers calculate means from five- or seven-point Likert scales under the assumption that the intervals are roughly equal. If you do this, note it as a methodological assumption in your analysis.
What is ratio data in a survey?
Ratio data is any measurement that has a true zero and equal intervals between values. In a survey context, this includes questions about age, income, number of purchases, hours spent on a task, or headcount. Because zero is a real, meaningful value, you can make ratio comparisons — for instance, a respondent who made 10 purchases bought exactly twice as many as one who made 5.
Is a Likert scale ordinal or interval?
A Likert scale is technically ordinal. The labels — such as "strongly disagree" to "strongly agree" — form a clear order, but the psychological distance between each step is not guaranteed to be equal. Many researchers treat Likert scales as interval data in practice, particularly with five or seven response options, because doing so enables richer statistical analysis. This is an accepted pragmatic choice, not a theoretical error, provided you acknowledge it.
Why does the level of measurement matter for statistical analysis?
Different statistical tests require different data types. If you try to calculate an average on nominal data — such as finding the "average country" — the result is meaningless. If you use only the median for ratio data, you discard useful information. Matching your analysis to the correct measurement level ensures your conclusions are valid and that you extract the most insight your data can support.
What is an example of interval data in a survey?
The most common example in surveys is a Likert-type rating scale, such as a 1–7 satisfaction scale where the gaps between points are treated as equal. Outside of surveys, temperature in Celsius or Fahrenheit is the standard textbook example — 0°C is not the absence of temperature. Calendar years are another: the year 0 does not mark the beginning of time. The defining feature is equal spacing without a meaningful absolute zero.