In 2022, we create an estimated 2.5 quintillion bytes of data every day. With such vast volumes of data readily available online, you can understand why businesses are so keen to invest time and money into harnessing it.
There are different types of data, which is commonly categorised into two types – structured and unstructured. Let’s explore the two:
Structured Data
Structured data is data that is in a predefined, tabular format. The rows and columns in tabular data often have relationships with one another. We call these relational databases.
| A | B | C | |
| 1 | Name | Address | Phone Number |
| 2 | Joe Bloggs | 3 Applewood Street | 01234567890 |
| 3 | Sarah Jones | 25 Loop Way | 02209876109 |
Also known as quantitative data, each data point has a numerical value associated with it. Structured data is organised and logical, which allows machine learning algorithms to decipher its contents easily.
To manage the structured data, Data Scientists use programming languages such as SQL (Structured Query Language). SQL allows Data Scientists to query a database to extract valuable information. Other tools used for structured data analysis are OLAP, SQLite, MySQL, and PostgreSQL.
Structured data is in industries such as eCommerce, banking and online booking platforms like booking.com.
Whilst structured data is useful in certain instances, it does have its limitations. Because structured data is predefined, it is only used for its intended purpose, limiting flexibility and usability. Structured data is also typically stored in data warehouses, where the schema of the data is rigid. If a change in the data is required, all structured data must be updated, which is time and resource-heavy.
Unstructured Data
Comparatively, unstructured data is not pre-defined. Some examples of such data are text, images, video and audio.
Also known as qualitative data, due to its non-numerical format. Because unstructured data comes in many forms and is disorganised, machine learning algorithms cannot process it using traditional methods.
To make use of this data, it is first cleaned and stored. Frameworks such as Hadoop are used, whilst MongoDB and Azure store the data to help to generate valuable insights. We can gather unstructured data through methods such as web scraping, APIs and surveys.
Estimates suggest that around 80%-90% of the world’s data is currently unstructured. This data is present in industries such as social media, surveillance and healthcare. Image processing, natural language processing and speech recognition typically use unstructured data.
The limitation of unstructured data is no more apparent than with the sheer size and scale of the data. Its lack of schema makes it expensive and difficult to store. NoSQL databases and data lakes store unstructured data natively. As the amount of unstructured data in any business grows, the storage capacity required becomes harder to attain. Specialised tools are also required to analyse this data, contributing to cost and time efficiency.
