Learn AI /Data and Infrastructure /Structured data
Structured data
- Starter
- Data and Infrastructure
- Programme & data teams
Also called: tabular data.
Back to the milestone you came from
In five seconds
What it is
Information organised into predefined fields, often in rows and columns, so software can process it predictably.
In plain words
It is information arranged the same way every time, usually in rows and columns, so software can handle it predictably.
An everyday example
A delivery service asks every sender to place the recipient, postcode, weight and service type in the same labelled boxes. Scanners can then check whether a field is missing, group parcels by postcode and route them automatically. If one sender writes the postcode under “weight” or uses a different label entirely, the system cannot process the parcel reliably. Structured data works in the same way: predictable fields make automated checking, sorting and matching possible.
Real structured data is not limited to paper-like cards. It can live in databases, update automatically and be linked across many tables. Also, some useful data is only partly regular, so the neat-card image can hide the fact that many real systems mix structured, semi-structured and unstructured information.
Remember thisStructured data puts each record into the same predefined fields, allowing systems to validate, search, aggregate and link records predictably.
Understand it
Understand it
Why it matters
Structured data affects what questions software can answer quickly and safely. If data is well organised, teams can compare like with like, automate routine reporting and feed tools with cleaner inputs. If it is badly structured, results can be misleading even when the chart or model looks polished. For non-specialists, this matters in procurement, oversight and daily operations: before trusting an output, ask whether the underlying fields are consistent, complete and current. Good structure does not guarantee truth, but poor structure often guarantees extra work and higher risk.
From records to usable analysis
Show how consistent rows and columns let software turn records into filters, counts and model-ready inputs.
01
Records collected
Start with several items to track, such as documents or projects. Each item becomes one row.
Draw a small table with 4 rows and faint placeholders, labelled 'one row = one item'.
02
Fields standardised
Each column has one meaning, such as title, country, date or budget. Formats are made consistent.
Show column headers in bold: title, country, year, topic. Add a small note: 'same field meaning in every row'.
03
Software operates on the table
Because the layout is regular, software can sort, filter, count and combine tables using shared fields.
Add icons or simple labels above the table: sort, filter, count, join. Draw arrows from the table to these labels.
04
Useful outputs
The same data can feed dashboards, reports, alerts or AI systems that expect clearly defined inputs.
Show four output boxes: chart, summary table, recommendation list, model input. Connect them from the operations stage.
A left-to-right flow with four panels. Read from raw records, to named columns, to software operations, to outputs. Each panel shows the same small dataset becoming easier to analyse because fields stay consistent across rows.
Hovering or tapping each stage highlights the relevant part of the table and shows one example question, such as 'count projects by region' or 'find documents in Arabic from 2022'.
Describe this without the diagram
The diagram explains why regular structure matters. First, several items are collected, with one row representing one item. Next, the rows are organised into columns with fixed meanings such as title, country, year and topic. Then software can perform operations on those columns, including sorting, filtering, counting and joining with another table using a shared field. Finally, these operations produce outputs such as charts, summary tables, recommendation lists or model-ready inputs. The key idea is that consistent fields make reliable processing possible.
With reduced motion: A static four-panel flow shows the same dataset moving from collected records, to standardised columns, to software operations, to outputs, without animation. Highlight colours and labels replace motion.
Structured data is information arranged in a fixed pattern. Usually that means rows and columns, like a spreadsheet or database table. Each row describes one item, and each column holds one type of value, such as date, country, budget or status.
What goes in? Clear fields and consistent formats. For example, a programme table might have one row per project and columns for project code, start date, region and spending. If dates use the same format and regions use the same names, the data is easier for software to work with.
What happens next? Software can search, sort, filter, join and count the data. "Join" means linking two tables using a shared field, such as project code. Because the structure is regular, a model or reporting tool can quickly answer questions like "Which projects in East Africa started in 2024?" or "What is the average spend by region?"
What comes out? Summaries, charts, alerts, recommendations and model inputs. Structured data is often easier to check and govern than free text, but it can still contain errors, missing values or outdated categories. It is different from unstructured data, such as long documents, images or audio, which do not come in neat columns by default.
How “A parcel label with fixed boxes” maps onto the real thing
| In the story | In the AI system |
|---|---|
| one index card per document | one row per record |
| same boxes on every card | the same columns or fields for each record |
| pulling cards by topic, language and year | filtering and querying structured data |
| inconsistent or blank boxes | noisy data or missing data |
A worked example
A programme officer in Kathmandu is preparing a side-event schedule for circulation before close of business. Proposals and available rooms are recorded in separate tables linked by a room ID.
Starting from: The proposal table has fixed fields for event ID, title, format, time slot, clearance status and room ID. The room table has fields for room ID, capacity and accessibility features.
- The system checks that every proposal has an event ID, a valid time slot and a clearance status.
- It filters for proposals that have been cleared and have an assigned room ID.
- It joins each proposal to the room table using the shared room ID field.
- It sorts the matched records by time slot and formats them for the schedule.
Result: The programme officer receives a consistent list of cleared events with the corresponding room details. The join works because both tables follow a stable schema and use the same room ID as a key.
In a United Nations workplace
Following an audit observation on inconsistent action tracking, the duty officer supporting a UN Country Team working group must consolidate the management response by close of business. The tracker uses one row per action, with fixed fields for action ID, responsible entity, due date, status and evidence link.
The predefined fields allow the duty officer to detect missing values, filter overdue actions and group them by responsible entity. The action ID also provides a stable key for linking each tracker row to its supporting evidence.
Before you try thisThe tracker and supporting evidence may contain confidential internal audit material. They must not be copied into an external or unapproved AI service. Keep them in approved systems, apply the required access controls and confirm that all entities use the same status definitions before aggregating results.
Where you may hear it
- A meeting on the response to an internal audit observation
- A vendor demo for a new reporting portal
- An inter-agency working group reviewing data standards
- A dashboard review before a progress report is cleared
- A database migration meeting with the ICT focal point
Avoid confusion
Avoid confusion
Often confused with
Common misconceptions
Common misconception: If information is in a spreadsheet, it is automatically high quality structured data.
Correction: A spreadsheet can still contain merged cells, mixed date formats, blank values, duplicate rows or columns that change meaning. The container is not the same as the quality of the structure.
Common misconception: Structured data is only numbers.
Correction: Structured data can include text categories too, such as country names, language codes or status labels. What makes it structured is the consistent field layout, not whether the values are numeric.
Common misconception: Structured data is always better than unstructured data.
Correction: It is easier for many systems to process, but it can leave out nuance found in documents, images or audio. Many real tasks need both: structured fields for consistency and unstructured material for context.
Quick check
Which example best fits structured data?
Structured data uses consistent fields for each record. The project table has a regular layout that software can sort, filter and analyse. The other examples may contain useful information, but not in a fixed column-based structure.
Go deeper
Go deeper
The simple version
It is information arranged the same way every time, usually in rows and columns, so software can handle it predictably.
The technical version
Structured data uses a predefined schema, meaning an agreed layout of fields, data types and relationships. In practice, this often appears as relational database tables or well-formed spreadsheet columns where each row is a record and each column represents one attribute with a consistent meaning. Because the schema is stable, systems can validate values, index fields for fast search, aggregate across records, and join tables through keys such as an ID code. This differs from unstructured data, where the useful information is embedded in free text, images or audio rather than placed into fixed fields from the start.
Reviewed external resources
Primary sourceAI Risk Management Framework (opens in a new tab)NIST
The NIST AI Risk Management Framework helps readers connect data quality and governance to AI risk, and suits practitioners who need a practical official framework.
Intermediatelink checked 2026-07-31
Primary sourceOECD AI Principles (opens in a new tab)OECD
The OECD AI Principles provide high-level guidance on trustworthy AI, including attention to data quality and governance, and suit policy and management audiences.
Intermediatelink checked 2026-07-31
EncyclopaediaSemi-structured data (opens in a new tab)Wikipedia
This Wikipedia page gives a basic introduction to semi-structured data, which helps readers compare it with structured data and suits beginners who want a quick overview.
Intermediatelink checked 2026-07-31
VideoVideo explanations: Structured data (opens in a new tab)YouTube search
This YouTube search results page helps readers find different video explanations of structured data, and suits people who prefer to learn by watching; it is a search page, not a single vetted lesson.
search results, not one pagelink checked 2026-07-31
DiscussionCommunity discussion: Structured data (opens in a new tab)Reddit search
This Reddit search results page lets readers see how people discuss and question structured data in practice, and suits readers who want informal examples; it is a search page, not a single authoritative source.
search results, not one pagelink checked 2026-07-31
Where this came from
Drafted with a language model through an offline pipeline, checked against this site's schema, and every external link above fetched to see whether it answers. Treat it as a starting point and check anything that matters against the primary sources. How this site is built and what that means.
The one thing to remember
Structured data puts each record into the same predefined fields, allowing systems to validate, search, aggregate and link records predictably.