What Are the Types of Data?
Root Concept
Data comes in three shapes — neat tables, labelled records without a fixed table, and raw files like photos and audio — and the shape decides how much work it takes before you can analyse it.
CodePLU Goal
Upgrading Human Mental Models
Learn how to think in Workflows
Concept Development By codeplu.com
Three types of data, and the everyday things that belong to each
Why Does the Shape of Data Matter?
Think about three ways of keeping track of your money. A bank statement is a neat table: every line has a date, a description, and an amount, always in that order. A shoebox of receipts is messier, but each receipt is still labelled — you can find the total and the shop name on every one, even though no two look alike. And a voicemail from your accountant is just sound: the information is in there, but nothing marks where the numbers are.
Data behaves in exactly these three shapes, and the names follow the neatness. Structured data sits in fixed rows and columns. Semi-structured data carries labels on each piece but does not force every record to match. Unstructured data is raw content — text, images, audio, video — with no labels at all.
This is not trivia. The shape decides how much work stands between you and an answer: a total from a table takes seconds, while the same total hidden across a thousand emails takes real effort. Most of the world's data is the messy kind, which is exactly why the skill of handling it is worth having. In the playground you will build data down into its three types, and each type down to something you have already seen in that shape.
How Do the Three Types Differ in Practice?
What makes data 'structured'?
Structured data lives in a fixed shape agreed in advance: named columns, one kind of value in each, and every row following the same pattern. A sales spreadsheet with date, item, quantity, and price is the classic case, and so is almost any database table behind a shop or a bank. Because the shape is known, a computer can total a column, sort by date, or filter a category without being told anything else — which is why answers from structured data arrive in seconds. The trade-off is rigidity: the moment you need to record something the columns did not anticipate, you are either adding a column for everyone or squeezing the information somewhere it does not belong.
What does 'semi-structured' actually mean?
It is the middle ground, and the name confuses people more than the idea deserves. Semi-structured data labels every piece of information — customer name, delivery address, item list — but does not insist that all records look alike. One online order might carry gift-wrapping details and a message; the next has neither, and that is fine, because nothing was promised in advance. This is how most information moves between apps and websites today. The advantage is flexibility: new kinds of information can appear without redesigning anything. The cost is that you cannot assume a field exists, so anything reading it has to cope with records that differ.
Why is unstructured data both the biggest pile and the hardest?
Unstructured data is content with no internal labels: the text of a complaint email, a product photo, a recorded call, a scanned invoice. Nothing inside marks which part is the customer name or the amount. It is thought to make up the large majority of everything organisations store, and until recently most of it simply sat there unused, because answering a question from it meant a person reading, watching, or listening. Modern AI changed the economics of exactly this — extracting the topic of ten thousand emails or reading the total from scanned receipts is now practical. That is a large part of why AI became commercially interesting so quickly.
How else can you split up data — numbers and categories?
Alongside the neatness split there is a second, equally useful one, and mixing them up causes real mistakes. Some values are numbers you can do arithmetic on, like price, age, or weight. Others are categories that merely name a group: country, product type, payment method. The trap is that categories are often stored as numbers — postcodes, customer IDs, a rating from one to five, or 1 and 2 standing for two branches. Nothing stops a computer averaging them, and the result is nonsense: the mean of your customer IDs means precisely nothing. Before calculating anything, check whether a number is a quantity or just a label wearing a number's clothing.
How do you turn messy data into something you can count?
In practice, most analysis ends up structured, because tables are what tools and statistics expect. So the work is extraction: pull the pieces you need out of the mess and put them in columns. From semi-structured orders you might lift customer, date, and total into a table. From unstructured complaint emails you might extract the product mentioned and whether the tone was angry, turning prose into two neat columns. Every extraction throws information away, and that is the point — you are deciding what matters for your question. Which means the same shoebox of emails can become several completely different tables depending on what you are trying to find out.
Real World Example
How Does One Clinic Hold All Three Types at Once?
Follow one patient visit through a small clinic's systems, and watch three different types of data pile up from a single event.
The appointment record is structured
One row in a table: patient number, date, time, doctor, appointment length, and whether the patient attended. Every appointment looks identical in shape, which is why the clinic can instantly answer 'how many no-shows did we have last month?'
The intake form is semi-structured
The online form saves labelled fields, but patients differ: some list three medications and others none, some add an allergy section, some skip half of it. Nothing forces the records to match, so the shape flexes from patient to patient while staying labelled throughout.
The consultation notes and scan are unstructured
The doctor's typed notes are prose. The X-ray is an image. Neither contains any marker saying 'this part is the diagnosis'. The information is genuinely there — often the most important information of the whole visit — but no simple query can reach it.
Which question you can answer depends on the type
'What is our average appointment length?' comes straight from the table in seconds. 'Which allergies are most common?' needs the semi-structured forms gathered into a column first. 'How often did notes mention back pain?' requires reading a thousand notes — or a tool that reads them for you.
The clinic's real project is extraction
To study back pain, someone must pull that concept out of the prose and into a yes-or-no column beside each appointment. Once the mess becomes a table, ordinary analysis works. That step — from unstructured to structured — is where most of the effort and most of the value sits.
FAQs
Final Words
Data arrives in three shapes: neat tables, labelled-but-flexible records, and raw content with no labels at all. In the playground you sorted a sales spreadsheet, a saved online order, and a pile of customer emails into their types — and the shape of each one is what decides how quickly you can get an answer out of it.
Two ideas are worth keeping. Most of the world's data is the messy kind, so the valuable skill is extraction: pulling what your question needs into columns you can count. And watch out for numbers that are really labels — averaging customer IDs is the classic way to produce a confident, meaningless result.
Continue This Track
This concept is part 2 of Data Science From the Ground Up.
What Is Data Science?
Data science is not just charts and not just coding. Learn what it really is by seeing how it differs from the two jobs sitting either side of it — and why it always ends in a decision.
What Are the Types of Data?
Some data arrives in tidy rows, some arrives with labels but no table, and some arrives as photos and voice notes. Learn to tell the three types apart and what each one costs you to use.
What Are the Steps of the Data Analysis Workflow?
Most beginners start with data and hope a chart shows something. Professionals start with a question. Walk the five steps of a real analysis, from the question to the answer someone acts on.
How Do You Clean Messy Data Before Analysing It?
Real data arrives broken: duplicates, blanks, mismatched formats, and impossible values. Learn what each mess does to your answer and the fix that clears it.
