Attribute, Ordinal and Numerical Data
You cannot choose an analysis until you know what your fields will actually hold. This page reproduces what the supplied teaching source says, sets out the decisions that must be made before collection opens, and marks the one place where its filing does not match the rest of the field.
The three shapes described in the supplied source
The teaching material introduces data shapes while explaining the difference between the two major research designs. It is not a taxonomy section, and it is not exhaustive — but the three shapes it does describe cover most of what a project research instrument will actually collect. Each of the terms below also appears in the consolidated research terminology glossary.
Definitions in the source's own terms
- Attribute data (also called categorical data)
- Data produced by classifying things into different categories, such as gender of students, nationalities or age. The source notes that for analysis we are often interested in the number of people in each category, and in the counts or proportions derived from them.
- Ordinal data
- Data represented in ordered categories. The source's example is a satisfaction scale on which the categories run from low levels of satisfaction to higher ones, and it notes that this order may be reflected in the analysis.
- Quantitative (numerical) data
- Information represented using a numerical scale of some sort which is based on some agreed measurement process — the source's examples of agreed scales are time, distance, height and weight.
Attribute data: classification into categories
Attribute data is what you get when you describe aspects of a situation by sorting things into categories. The source's illustrations are demographic — gender of students, nationalities, age — and the categories carry no order. One nationality is not more than another.
The source then makes a move that is easy to skim past. Having called this qualitative description, it says that for the sake of analysis we are often interested in the number of people in each category, and that these counts, or the proportions derived from them, are frequently what we want to represent or analyse.
So the data is categorical, and the analysis is arithmetic. That is worth naming plainly, because it is the seed of the classification problem set out further down this page.
One consequence follows immediately. Categories have to be decided before collection, because you cannot count into a category you did not create. The source's description of quantitative collection uses the same idea from the other direction, describing structured instruments that fit diverse experiences into predetermined categories — a point that becomes sharper alongside the thin treatment of data collection methods elsewhere in the material.
Ordinal data: categories that carry an order
Sometimes the categories have a sequence. The source's worked illustration is a satisfaction question: students are asked to indicate how satisfied they are with a service, and respond on a five- or seven-point scale running from very dissatisfied to very satisfied.
Note the internal logic: ordinal data is presented as attribute data with an order added. The order is the only difference, and the source says it may be reflected in the analysis — not that it must be, and not how.
Numerical data: measurement against an agreed scale
The third shape is the one the source treats as fully quantitative. Its defining feature is not that the values are numbers — counts of category members are numbers too — but that the numbers come from an agreed measurement process. Time, distance, height and weight are the source's examples of entities with standard numerical scales already defined for them.
The claim the source makes for this shape is about resolution. Quantitative data usually provides a much more sensitive description of a phenomenon than simply being able to categorise the information, and the patterns visible in it are often very informative.
That word sensitive is doing real work. A category tells you which bucket something is in; a measurement tells you how far it moved and by how much it differs from the next case. Where a delivery decision turns on magnitude rather than membership, this is the shape you need.
WHAT THE SOURCE ATTACHES TO EACH SHAPE
| Shape | How it is produced | What the source says you do with it | Classified in the source as |
|---|---|---|---|
| Attribute or categorical | Classifying things into different categories that have no inherent order | Count members per category; derive proportions; use those as the basis for analysis | Qualitative |
| Ordinal | Classifying into categories that run in a sequence, such as a satisfaction scale | The ordering may be reflected in the analysis. No technique is specified | Qualitative |
| Numerical | Measurement against a numerical scale based on an agreed measurement process | Summarise, compare and generalise; test hypotheses derived from theory; estimate the size of a phenomenon | Quantitative |
The classification column reports the supplied source's own filing, which is contested. See the next section.
The classification tension the source leaves open
The source states that the more common forms of data described as qualitative are attribute or ordinal data. Read strictly, that puts a table of counts and a rating scale on the qualitative side of a distinction the same passage defines as the difference between interpretation and counting.
What the supplied source asserts
- Qualitative data is non-numerical and represents different qualities of a phenomenon.
- The more common forms of data described as qualitative take the form of attribute or ordinal data.
- Counts of category members and proportions derived from them are often the basis of the analysis.
- Quantitative data uses a numerical scale based on an agreed measurement process.
What the supplied source does not settle
- Whether counting members of a category is a qualitative or a quantitative act.
- How ordered categories should be analysed, or whether their intervals may be treated as equal.
- Where a rating scale belongs when the analysis is entirely arithmetic.
- Any alternative framework for classifying data, including levels of measurement.
The practical resolution is to stop treating the label as load-bearing. The same tension runs through the broader distinction covered in qualitative and quantitative research compared, and in both places the useful discipline is to describe the operation rather than argue about the category.
Deciding the shape before you build the instrument
Data shape is a design decision, not a reporting decision. Once responses are in, the shape is fixed — you can always collapse a measurement into categories, but you can never recover a measurement from a tick box. The sequence below is offered as practice; the supplied source does not set out a procedure of this kind.
Fixing the shape of each field
Write the sentence you want to be able to publish
Not the question — the finding. “Forty-one per cent of respondents in delivery roles reported…” and “Median approval time fell by eleven days…” are different sentences requiring different data shapes.
Decide whether the answer is a membership, an order or a magnitude
Membership gives attribute data. A position on a ranked set of options gives ordinal data. A quantity measured against an agreed scale gives numerical data.
Fix the categories in advance where the answer is a membership
Closed lists must be complete and mutually exclusive before collection begins, because counts cannot be made into categories that were never offered.
Choose the response format that yields the shape
An open text box does not produce a count. A five-point scale does not produce a measurement. Match the format to the sentence from step one.
Confirm the intended analysis is supported
State the analysis for each field before collection. If the technique you have in mind assumes equal intervals and the field is ordered categories, resolve that now, with a citation.
Record the shape of every field in your data dictionary
One line per field: name, shape, permitted values, intended analysis. This is the artefact that keeps your methodology chapter honest six months later.
Reading data shape in familiar project artefacts
Most project managers already handle all three shapes daily without naming them. Recognising which is which in your existing reporting is the quickest way to build the habit, and it tells you which of your organisation's data can be reused as evidence — an issue taken further in primary and secondary data sources.
SHAPES IN EVERYDAY DELIVERY DATA — AN APPLIED READING, NOT A SOURCE TABLE
| Artefact or field | Shape | Why |
|---|---|---|
| Contract type recorded against each package | Attribute | A closed list with no inherent order; reported as counts and proportions |
| Risk consequence rating on a register | Ordinal | Ordered categories from low to high; the order is meaningful, the spacing is not stated |
| Days between issue raised and issue closed | Numerical | Measured against an agreed scale for time |
| Stakeholder sentiment captured on a rating scale | Ordinal | Ordered response categories, exactly the shape of the source's satisfaction example |
| Cost variance against baseline in dollars | Numerical | Measured against an agreed monetary scale |
| Whether a gate review was held | Attribute | Two categories, no order; the analysis is a count |
This mapping applies the source's three definitions to common delivery artefacts. The artefacts themselves do not appear in the supplied teaching material.
Where this goes wrong
What to carry forward
- Three shapes are described in the supplied source: attribute or categorical, ordinal, and numerical against an agreed measurement scale.
- The source files attribute and ordinal data under qualitative data. Many texts do not, and this library reports that as an unresolved tension rather than picking a side.
- The source's own attribute discussion is about counts and proportions, which is the strongest reason to treat the label as unhelpful.
- Ordered categories are ordered; nothing in the source says their intervals are equal or licenses any particular arithmetic on them.
- Numerical data is defined by the agreed measurement process behind it, and is claimed to give a more sensitive description than categorisation.
- Fix the shape of every field before collection, and record it. You can collapse detail later, but you cannot recover it.
Frequently asked questions
Does the supplied source really call a rating scale qualitative?
Yes. It states that the more common forms of data described as qualitative take the form of attribute or ordinal data, and its worked ordinal example is a five- or seven-point satisfaction scale. Many research methods texts classify that data as a level of measurement handled quantitatively. This library reports the source's framing as the source's and flags the disagreement rather than resolving it.
Can I calculate an average of an ordinal scale?
The supplied source does not say. It states only that the categories are ordered and that this order may be reflected in the analysis; it makes no claim about equal intervals and prescribes no technique. If you intend to average ordered categories, justify it from a text you cite and record the decision in your methodology chapter.
What makes data numerical rather than just numbers?
In this source, an agreed measurement process. Counts of category members are numbers but are discussed under categorical data; numerical data means values on a scale that already exists by agreement, such as time, distance, height or weight. The distinction is about where the scale comes from, not about whether digits appear.
How many points should my rating scale have?
The supplied material does not recommend a number. Five and seven appear only inside an illustrative example about student satisfaction, and there is no discussion of midpoints, labelling or scale length. Any choice you make needs a survey design source of its own.
Which shape should I collect if I am not sure?
Where a genuine measurement is available, collect it. A measurement can always be collapsed into categories at analysis time, whereas a category can never be expanded back into a measurement. The source does not state this rule, but it follows directly from its own definitions.
References and source attribution
- Bryman, A. 2016, Social Research Methods, 5th ed., Oxford University Press, Oxford.
- O'Leary, Z. 2017, The Essential Guide to Doing Your Research Project, 3rd ed., Sage Publications, London.
- Naoum, S. G. 2013, Dissertation Research & Writing for Construction Students, 3rd ed., Routledge.
- The supplied teaching source: teaching notes and slide decks on developing a research topic, section 5, 'Design your research', qualitative and quantitative data passages. Institution, unit code and authors withheld.
Suggested questions for Ask KEVOS
- Classify each field in my draft survey as attribute, ordinal or numerical data.
- Show me the exact wording the supplied source uses about attribute and ordinal data.
- Explain the classification tension between the source's framing and the levels of measurement view.
- What can I legitimately claim from ordered category data according to this source?
- Draft a data dictionary structure for a research instrument on procurement performance.
- Which of my existing project reports already contain numerical data I could reuse as evidence?