Structured Data and Data Models#

📦 Data Preparation 🧬 Data Types & Structure Lesson 004

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

How organised is the data?#

Beyond individual value types, data has an overall degree of structure — how regularly it is organised — and this determines which tools can work with it and how much preparation it needs before analysis. The standard three-way split is one of the most useful classifications in the field.

Structured, semi-structured, unstructured#

Structured data is organised into a defined model — rows and columns, with a fixed schema saying what each field is. A database table or a tidy spreadsheet is structured: every record has the same fields, each of a known type. Structured data is the easiest to query, aggregate, and analyse, and it is where most of this course operates.
Semi-structured data has organisational markers but no rigid table schema — tags or keys that give it shape while allowing records to vary. JSON and XML are the classic examples: hierarchical, self-describing, but not a fixed grid. Common from web APIs, it usually needs reshaping into structured form before tabular analysis.
Unstructured data has no predefined model: free text, images, audio, video. It is the most abundant kind of data in the world and the richest, but also the hardest to analyse — extracting structure from it (turning reviews into themes, images into labels) is often a project in itself.

The progression structured → semi-structured → unstructured runs from easiest to analyse, least abundant toward hardest to analyse, most abundant — which is much of why data work is challenging.

What a data model is#

A data model is the description of how data is organised: what entities exist, what attributes each has, and how they relate. Structured data has an explicit model (a table’s columns and types, a database’s schema); giving unstructured data a model is precisely the act of structuring it. Models are usually described at three levels of detail — a high-level conceptual view of the main entities, a logical view specifying attributes and relationships, and a concrete physical view of how it is actually stored — moving from what the business cares about down to how the database holds it.

Why structure and models matter for preparation#

The degree of structure sets the preparation workload. Structured data is close to analysis-ready and needs mainly cleaning. Semi-structured data needs reshaping into tables first. Unstructured data needs structuring — often the largest task of all. Recognising which kind you face, early in the Prepare phase, tells you how much work stands between the raw data and a valid analysis — and whether the question is even answerable with the effort available.

The caveat#

The categories are a spectrum, not sharp bins: a spreadsheet full of free-text notes is nominally structured but practically holds unstructured content in its cells, and “structured” guarantees a shape, not quality — perfectly structured data can still be riddled with errors, which is exactly why cleaning is its own section. Structure tells you how the data is organised, never whether it is right. The following lessons move from these broad kinds down to the concrete shapes data takes in spreadsheets and tables.

Hint