What Is Metadata? Definition, Types, and Examples

Metadata is data about data: information that describes another piece of data so people and systems can find it, understand it, trust it, and use it correctly. A photo's file holds the pixels; its metadata records when and where the photo was taken and with which camera. A database table holds rows of orders; its metadata records what each column means, where the data came from, who owns it, and when it was last refreshed.

Metadata is easy to overlook because it is rarely the thing anyone came for. Yet almost every useful action on data depends on it. Search engines rank pages using their titles and descriptions, music apps sort songs by artist and album, and an analyst can only pick the right revenue table if something tells them which one is the approved source. As organizations hand more questions to AI systems, metadata has become the context those systems need to answer correctly.

What Is Metadata?

The shortest definition is "data about data." A fuller definition comes from NISO's primer Understanding Metadata, which describes metadata as the information we create, store, and share to describe things, so that we can interact with those things and obtain the knowledge we need.

Three properties make something metadata rather than data:

• It points at something else. Metadata always describes a resource: a file, a web page, a book, a table, a dashboard, a model.
• It answers questions about that resource. What is it, who made it, when, where is it stored, what does it mean, can I use it?
• It depends on the role, not the content. A date is data in an orders table and metadata when it records when that table was last refreshed. The same value can be either, depending on what it describes.

Metadata Examples

Metadata is everywhere, and most of it is created automatically. The table shows the same pattern across common resources: the content itself on one side, the information about it on the other.

The communication row is a reminder that metadata can be sensitive on its own. Who called whom, when, and for how long can reveal a great deal without a single word of the conversation. Photo location data works the same way, which is why many apps strip it before an image is shared.

Types of Metadata

There are two common ways to sort metadata. The first comes from libraries and archives and applies to any kind of resource. NISO's primer groups it into three main types:

• Descriptive metadata helps people find and understand a resource: title, author, subject, keywords, abstract.
• Structural metadata describes how the parts of a resource relate to each other: the chapters of a book, the pages of a scanned document, the tables and keys of a database.
• Administrative metadata is what you need to manage a resource. It is often split further into technical metadata (file format, size, encoding), preservation metadata (what keeps it usable over time), and rights metadata (who may use it and how).

Data teams usually use a second grouping, built around what each kind of metadata helps them do:

• Technical metadata describes how data is stored and structured: databases, schemas, tables, columns, data types, keys, file locations, and row counts. Most of it can be harvested automatically from source systems.
• Business metadata describes what data means to the organization: definitions in a business glossary, the owning team, the domain or department, classification such as personal or confidential, and approved uses. It usually needs people to write and maintain it.
• Operational metadata describes what happens to data: when it was last loaded, which pipeline produced it, whether quality checks passed, how often it is queried, and its lineage from source to report.

The two groupings overlap rather than compete. Technical metadata in the data team's sense is part of administrative metadata in the library sense, and business definitions are a form of descriptive metadata. What matters is that each kind has an owner and a home, because they are captured in very different ways: technical metadata by scanning systems, business metadata by people who know the business, operational metadata by the pipelines and tools that touch the data. When metadata is collected continuously and used to drive actions, such as alerts or access decisions, it is often called active metadata.

Metadata vs Data

The difference is one of role. Data is the content people want to use; metadata is the description that makes it usable. A few practical consequences follow:

• Metadata is much smaller than the data it describes, which is why a data catalog can index an entire data estate without copying the data itself.
• Metadata outlives copies. When a table is copied, renamed, or moved, its technical metadata can be harvested again, but business definitions and ownership are lost unless they are kept somewhere central.
• Metadata can carry risk. As the examples show, metadata about people, such as call records or photo locations, can be personal information in its own right.
• Bad metadata misleads quietly. A wrong number in a report is often caught; a wrong description of which table holds approved revenue can send every analyst to the wrong source without anyone noticing.

Metadata Standards

Metadata is only useful if systems and people describe things the same way. Standards make that possible, and a few cover most of what organizations meet:

• Dublin Core. Fifteen general-purpose elements, such as Title, Creator, Subject, Date, Format, and Rights, used widely for documents and web resources. It is published as ISO 15836.
• Exif. The standard that digital cameras and phones use to embed technical metadata in image files, including camera settings, timestamps, and location. It is maintained by the Japanese camera industry association CIPA.
• ISO/IEC 11179. The international standard for metadata registries, which describes how to define and register data elements so that the same concept, such as a customer identifier, is described consistently across systems. The framework part was most recently revised in 2023.
• Schema.org. A shared vocabulary for structured data on web pages that search engines use to understand what a page is about.

Inside a company, the business glossary plays the same role for business metadata: it is the internal standard that says what "customer," "active user," or "net revenue" means.

Why Metadata Matters for AI

An AI assistant or agent that answers questions about company data does not see the business the way an employee does. It sees table names, column names, and values. Metadata is how it learns the rest:

• Meaning. Business definitions tell a model that net_rev is revenue after refunds and which table is the approved source for it, so it does not guess from a column name.
• Trust. Operational metadata such as freshness, quality results, and lineage tells it whether a dataset is current and fit to answer from.
• Permission. Classification and ownership tell it which data contains personal or confidential information and who may see it.

Without that context, AI answers are fluent but unreliable, which is the gap a metadata layer for AI is meant to close. It is also why AI-ready data is mostly a metadata question: the data often exists already, while the descriptions that make it usable do not.

From Metadata to Metadata Management

Every organization already has metadata. The problem is that it is scattered: technical metadata sits in each database, definitions sit in spreadsheets and people's heads, and lineage sits in pipeline code. Metadata management is the practice of collecting it in one place, keeping it current, and assigning people to own it. The usual building blocks are a data catalog for the inventory, a business glossary for definitions, and data lineage for flows.

Dawiso is built for that job. Its data catalog harvests technical metadata from connected sources, the business glossary holds the business definitions and owners, and interactive lineage shows how data moves from source to report. Together they give people, and the AI systems working alongside them, one place to look up what data exists, what it means, and whether to trust it.

Conclusion

Metadata is the description that turns stored values into usable information. It tells a photo app where a picture was taken, a search engine what a page is about, and an analyst which table holds the approved revenue figure. The classic types, descriptive, structural, and administrative, apply to any resource; data teams add the practical split between technical, business, and operational metadata. As more questions are answered by AI rather than by people who know the business, the quality of that description decides the quality of the answer.

Sources

• National Information Standards Organization (NISO) - Jenn Riley, Understanding Metadata: What Is Metadata, and What Is It For? (2017)
• Dublin Core Metadata Initiative - Dublin Core Metadata Element Set, Version 1.1 (ISO 15836)
• Camera and Imaging Products Association (CIPA) - CIPA standards, including DC-008 Exchangeable image file format (Exif)
• International Electrotechnical Commission - ISO/IEC 11179-1:2023, Information technology - Metadata registries (MDR) - Part 1: Framework
• Schema.org - Schema.org vocabulary for structured data