Dark logo

The Three Types of Data

Published March 13, 2017
Doug Rose
Author | Agility | Artificial Intelligence | Data Ethics

When organizations capture and analyze big data to extract knowledge and insight from it, they often must aggregate three types of data:

  • Structured
  • Semi-structured
  • Unstructured

In this article, I highlight the differences among these three data types.

Structured Data

Comprising about 10 percent of all available data, structured data is typically stored in a relational database consisting of interrelated tables. Every record has a unique identifier (a primary key), and every data entry is “labeled” with a field name to simplify the process of using and managing the data. Structured data is like the bricks and mortar of the database world. It’s cheap, inflexible, and requires a lot of upfront development.

A good example of structured data is a typical office spreadsheet. When you fill rows with data, you have to stick to a pretty rigid format and structure. For example, suppose you have a column called “Purchase Date.” Each entry in that column must be in a specific format. You can’t have “Tuesday” as one entry and “March” as another. For a purchase date, you want to specify the day, month, and year, and you want all the entries to be consistent. One way to ensure consistency is to create a rule that every entry in the Purchase Date column must be in the MM/DD/YYYY format.

Without a consistent format, you would have a difficult time extracting specific data and using it to create a report, for example. Imagine trying to create a sales report for a certain date range if some dates were spelled out (such as February 2, 2019) while others were numerical (such as 03/29/2020) and some were abbreviated (such as Jan 23, 2020). Any query used to extract date information would need to cover all the variations in the date format.

Most data that an organization generates and uses internally is transaction data, which can be highly structured. Sources of structured data include the following:

  • Sales transactions
  • Inventory records
  • Flight and room reservation systems
  • Customer relationship management software
  • Financial records

With structured data, you have all your data neatly arranged in a relational database, where everything is labeled and organized. You know where everything is, and you know exactly where to find it. It’s like storing spices in labeled jars arranged alphabetically in a spice rack. You know where everything is and you know exactly where to find it.

Semi-Structured Data

Semi-structured data contains tags or other markers to “label” data entries without forcing the data into a strict structure such as a table. Semi-structured data is typically stored in a text file that includes metadata— the tags or markers that identify or provide additional information about the data entries. Files that include semi-structured data are commonly saved in one of the following formats, which are generally easy for humans to read and understand:

  • JavaScript Object Notation (JSON)
  • eXtensible Markup Language (XML)
  • Comma separated values (CSV)
  • Tab-delimited files

Sources of semi-structured data include the following:

  • Websites
  • Internet of Things (IoT) devices
  • Apps
  • GPS trackers

Unstructured Data

The most common type of data is everything that isn’t structured or semi-structured: it is unstructured data. Some analysts estimate that 80 percent of all data is unstructured. When you think about it, this makes a lot of sense. Think about the data you encounter every day:

  • Online images
  • Online video
  • Online audio
  • Email messages
  • Social media posts
  • Chat transcripts
  • PDFs

What does all this data have in common? Not much, and that’s part of the problem. None of this data is structured in any uniform way. The data contained in these files are not arranged in tables or entered into specific fields. The content may or may not be tagged to describe what it contains. Even the file formats differ. If you had a dozen documents you couldn’t open, you’d struggle to figure out what was in each one. The challenge would be even greater if you needed to extract details from the documents and aggregate relevant data.

Fortunately, companies such as Google have been working for years to overcome these limitations. Every time you search Google, Bing, or Yahoo!, you reap the fruits of their labor. Whenever search for something on these sites, you’re presented with a long list of links to a variety of content, including web pages, Word documents, PDFs, images, audio, and video. Organizations in diverse industries use many of the same and similar technologies to capture, store, retrieve, aggregate, and analyze unstructured data.

Putting Your Data to Work

Suppose you own a business that sells running shoes online and you want to identify your best customers. First, you may want to define what you mean by “best customer.” Is your best customer the one who spends the most money in your webstore? Is it the person who suggests ideas for improving your products? Is it the customer who recommends your product the most on social media? Maybe your best customer meets all these criteria.

If your definition of best customer is the person who spends the most in your webstore, all you need is a simple query to your transactional database to rank customers in order from those who spent the most to those who spent the least. For a more sophisticated 360-degree perspective, you could aggregate data from your transactional database, from social media sites such as Facebook and Twitter, and from other sources. Using customer names and email addresses you already have, you can crawl social sites to find out what customers posted about your product — maybe a comment on Facebook or Twitter or a video on YouTube. You can then analyze the aggregated data (structured, semi-structured, and unstructured) to identify patterns that are characteristic of a “best customer.”

As time goes on, you can capture more and more of your customers’ unstructured data, which will allow you to ask more sophisticated questions about your customers. For example, you may want to know which races they run in, so you can sponsor those races. Or you may wonder what other running products they purchase, so you can figure out how to expand your product selection. Answers to each of these questions can help you connect with your customers and sell more products.

As you begin to work with data, realize that all types of data have value, whether structured, semi-structured, or unstructured. Part of your job is to figure out ways to extract value from data regardless of how structured or unstructured that data may be.

Related Posts
November 6, 2017
Data Storytelling Metaphors

Data storytelling metaphors are a powerful tool to help your data science team tell stories. Data visualizations are not the best way to communicate complex problems.

Read More
August 5, 2021
Data Warehouses and Data Lakes

The key difference between a data warehouse and a data lake is how the database is structured and how it's used.

Read More
August 5, 2021
What’s the Difference between Data Mining vs Machine Learning?

In data science the difference between data mining vs machine learning is a key concept.

Read More
1 2 3 19
9450 SW Gemini Drive #32865
Beaverton, Oregon, 97008-7105
Dark logo
© 2022 Doug Enterprises, LLC All Rights Reserved
linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram