Types of Data: 11 Types, How to Set Them Up and How to Maintain Them

Types of data refers to the categories that describe what a value represents and how a system can store and process it. Data is classified in 2 ways. The first way is by what the data measures: qualitative data (nominal and ordinal) and quantitative data (discrete, continuous, interval and ratio). The second way is by how the data is organised: structured, semi-structured and unstructured data.

The 11 types of data each have a definition, a place of use, a difference from the neighbouring type, and steps to set up and maintain the type in a database or spreadsheet.

Key takeaways

• There are 11 types of data in this classification. The types of data decide what a system can do with a value, because each data type supports a different set of valid operations.
• According to Harvard University research from the Psycho-Acoustic Laboratory, 7 June 1946, measurement uses 4 scales: nominal, ordinal, interval and ratio.
• Structured, semi-structured and unstructured data describe how data is organised, and not what it measures.
• Setup cost depends on data volume, sampling frequency and the maintenance skills available.
• Choosing among the types of data starts with the question the data answers.

What are the 11 types of data?

There are 11 types of data, and they answer two questions. Eight types show what a value measures: qualitative, nominal, ordinal, quantitative, discrete, continuous, interval and ratio data. Three data types show how the data is organised: structured, semi-structured and unstructured data. Each of the 11 types of data has a one-line definition:

• Qualitative (categorical) data: descriptive labels and opinions.
• Nominal data: unranked categories.
• Ordinal data: ranked categories.
• Quantitative (numerical) data: measured or counted values.
• Discrete data: countable whole numbers.
• Continuous data: measured values with fractions.
• Interval data: equal steps with an arbitrary zero.
• Ratio data: equal steps with an absolute zero.
• Structured data: rows and columns under a fixed schema.
• Semi-structured data: tagged data without a rigid schema.
• Unstructured data: files with no predefined model.

Nominal and ordinal data are kinds of qualitative data. Discrete, continuous, interval and ratio data are kinds of quantitative data. Structured, semi-structured and unstructured data form a separate group, because they classify the organisation of data.

Why do businesses need types of data?

Businesses need types of data because a system can only validate, store and calculate a value when it knows what kind of value it is. Without that classification, a postcode is averaged like a price, a rating is subtracted like a distance, and reports disagree. The article What is Data Monetization? shows how data assets of these types are turned into financial value.

According to MIT Sloan School of Management research from the Center for Information Systems Research (CISR), 20 May 2021, most company data has low data liquidity, because it is trapped in local processes, locked in closed platforms, duplicated or poorly defined. Clear types, units and formats address the "poorly defined" part directly. The article on data liquidity explains why shared definitions make data easier to reuse and recombine.

How are the types of data grouped?

The types of data are grouped by measurement and by organisation. Measurement splits the types of data into qualitative and quantitative data, and organisation splits them into structured, semi-structured and unstructured data. The table below lists each type with its group, an example and what it supports:

Evidence from measurement theory explains the order of the first four rows. Each scale builds on the scale below it. The original paper is "On the Theory of Scales of Measurement" by S. S. Stevens in Science.

What are the qualitative types of data?

Qualitative types of data describe qualities and categories, and not quantities. The 3 qualitative types are qualitative (categorical) data, nominal data and ordinal data. Qualitative data is the broad class, nominal data covers categories without order, and ordinal data covers categories with a fixed rank.

Qualitative (categorical) data

Qualitative data is non-numerical, descriptive data that records characteristics, labels and opinions. Businesses use it for customer feedback, sentiment analysis and demographic grouping. Qualitative data differs from numerical data because it relies on language and tagging, and not on arithmetic. Examples are survey comments and product reviews.

Setting up qualitative data

To set up qualitative data, use a controlled list of tags for each descriptive field and a free-text field for open answers. Cost rises with the interpretation method: manual coding costs analyst hours, and automated language processing costs software or API fees.

Maintaining qualitative data

To maintain qualitative data, audit the labels periodically. The audit corrects inconsistent tagging and drift in the meaning of terms.

Nominal data

Nominal data is a set of unranked, mutually exclusive categories that serve only as labels. Examples are country, industry sector and product colour. Nominal data differs from ordinal data because no category carries more weight than another. A record belongs to exactly one category, and the labels carry no numerical value. Nominal data needs only one label column per record.

Setting up nominal data

To set up nominal data, create a reference table with a unique constraint, and point the main table to it with a foreign key.

Maintaining nominal data

To maintain nominal data, merge duplicate categories, such as "USA" and "United States".

Ordinal data

Ordinal data is a set of categories with a fixed order but no measurable distance between values. A survey scale from poor to excellent is the common example. Ordinal data differs from nominal data because the order is mandatory, and it differs from quantitative data because "good" minus "average" has no numerical answer. Among the qualitative types of data, ordinal data is the only one with a built-in order.

Setting up ordinal data

To set up ordinal data, store a rank number next to each label and record the scale in a data dictionary.

Maintaining ordinal data

To maintain ordinal data, review the scale when its meaning changes and validate that new entries follow the rank order.

What are the quantitative types of data?

Quantitative types of data are numbers that can be measured, counted and calculated. The 5 quantitative types are quantitative (numerical) data, discrete data, continuous data, interval data and ratio data. Discrete and continuous data differ by how values are produced, and interval and ratio data differ by the meaning of zero.

Quantitative (numerical) data

Quantitative data is information expressed as numbers that support arithmetic. Personal budgeting apps, financial forecasting and resource planning all rest on it. Quantitative data differs from categorical data because a system can calculate averages and differences from it. Examples include prices, quantities, durations and temperatures.

Setting up quantitative data

To set up quantitative data for a study, project or business analysis, structure the collection method so that every data point can be turned into a clean, measurable number. The PostgreSQL documentation lists exact numeric types and approximate floating-point types, and exact types suit currency.

Maintaining quantitative data

To maintain quantitative data, put processes in place that protect its accuracy, security and formatting over time. Reconciliation checks compare stored totals with their source records. Without maintenance, numeric datasets suffer from data rot, formatting shifts and accidental deletions.

Discrete data

Discrete data is a set of countable whole numbers that cannot be split into fractions. Page views, headcount and inventory units are examples. Discrete data differs from continuous data because no value can exist between two neighbouring whole numbers. Counting produces discrete data, and measuring with an instrument usually produces continuous data.

Setting up discrete data

To set up discrete data, use an integer field with a check constraint that rejects decimals, and negatives where they make no sense.

Maintaining discrete data

To maintain discrete data, run constraint checks and reclassify the field when the counted entity changes.

Continuous data

Continuous data is a measured value that can take any figure within a range, including fractions. Temperature, fluid pressure and a run time are examples. Continuous data differs from discrete data because it is measured with an instrument and not counted. Readings arrive as a stream of values over time.

Setting up continuous data

To set up continuous data, use time-series storage that records the timestamp, the unit and the sampling interval. Cost rises with sampling frequency and retention period.

Maintaining continuous data

To maintain continuous data, down-sample and archive old high-resolution readings, and recalibrate the sensors.

Interval data

Interval data is numerical data with equal distances between values and an arbitrary zero. Temperature in degrees Celsius and calendar dates are examples. Interval data differs from ratio data because a zero reading does not mean the absence of the property, so 40 degrees is not twice as hot as 20 degrees.

Setting up interval data

To set up interval data, store the unit with each value and store timestamps in UTC.

Maintaining interval data

To maintain interval data, check time-zone alignment and sensor drift.

Ratio data

Ratio data is numerical data with equal distances between values and an absolute zero. Weight, revenue and bank balance are examples. Ratio data differs from interval data because ratios are meaningful: a company with two million in revenue has twice the revenue of a company with one million.

Setting up ratio data

To set up ratio data, add a check constraint where zero means none, such as quantities and weights.

Maintaining ratio data

To maintain ratio data, audit the fields for negative values and reconcile totals against the ledger or source system.

What are the structured types of data?

The structured types of data describe how data is organised, from a fixed schema to no schema at all. The 3 types are structured data, semi-structured data and unstructured data. In most cases, a stricter schema makes queries faster and makes changes to fields harder.

Structured data

Structured data is data that follows a predefined model, usually rows and columns in a relational database. A customer address book and a table of transactions are examples. Structured data differs from the other 2 types of data because the schema is defined before any record is written.

Setting up structured data

To set up structured data, define tables, keys, field types and indexes. Cost depends on server capacity, licensing and the number of concurrent writes.

Maintaining structured data

To maintain structured data, rebuild indexes, run schema migrations and tune queries. The article A System That Lasts describes the same principle for keeping a system understandable as it grows.

Semi-structured data

Semi-structured data is data that carries tags or a hierarchy but no rigid relational schema. JSON from a web API and XML files are examples. Semi-structured data differs from structured data because different records in one dataset can hold different fields. Each record describes itself through its tags.

Setting up semi-structured data

To set up semi-structured data, use a document database or a JSON column, with schema validation on write.

Maintaining semi-structured data

To maintain semi-structured data, validate incoming payloads and remove unused fields, so that nested documents stay navigable.

Unstructured data

Unstructured data is data with no predefined model, such as text documents, audio, video, images and PDFs. Unstructured data differs from the other 2 types of data because unstructured data needs extraction pipelines, or machine learning models, before it can be queried. Document archives and media libraries store unstructured data.

Setting up unstructured data

To set up unstructured data, use object storage with a metadata catalogue so that files can be found. Cost scales with total volume and with the compute used for extraction.

Maintaining unstructured data

To maintain unstructured data, apply lifecycle policies that move rarely used files to cheaper archive storage, and re-run extraction when the models change.

How do you choose the right types of data?

To choose the right types of data, start with the question the data answers, and then match the type to the field. In most cases, deciding the type before collection avoids later rework. Choosing the right types of data follows 6 steps:

1. Define the question or report that the data serves.
2. Identify what each field measures, and place it on the scale from nominal to ratio.
3. Choose the organisation of the data by how stable the fields are: structured for stable fields, semi-structured for changing fields, unstructured for files.
4. Check the privacy and security rules that apply, such as the General Data Protection Regulation in the European Union or the HIPAA rules for health information in the United States.
5. Match the design to the skills of the team that maintains it, because an advanced time-series database on a team that only runs overnight batches leads to failures and cost overruns.
6. Record each field's type, unit and allowed values in a data dictionary.

What are the limits of this classification?

The classification of the types of data has 3 limits. The first limit is overlap: one field can belong to several types, such as a structured, ratio-scale field. The second limit is dispute: researchers have criticised the four measurement scales, and the scheme is a practical guide and not a complete theory. The third limit is scope: the setup and maintenance guidance in this article is general reasoning, and exact costs depend on the platform, the data volume and the team. Within those limits, the practical rule for the types of data is to classify each field by what it measures and how it is organised before collecting it.