Structured data vs. unstructured data: structured data is comprised of clearly defined data types whose pattern makes them easily searchable; while unstructured data – “everything else” – is comprised of data that is usually not as easily searchable, including formats like audio, video, and social media postings.
Unstructured data vs. structured data does not denote any real conflict between the two. Customers select one or the other not based on their data structure, but on the applications that use them: relational databases for structured, and most any other type of application for unstructured data.
However, there is a growing tension between the ease of analysis on structured data versus more challenging analysis on unstructured data. Structured data analytics is a mature process and technology. Unstructured data analytics is a nascent industry with a lot of new investment into R&D, but is not a mature technology. The structured data vs. unstructured data issue within corporations is deciding if they should invest in analytics for unstructured data, and if it is possible to aggregate the two into better business intelligence.
What is Structured Data?
Structured data usually resides in relational databases (RDBMS). Fields store length-delineated data phone numbers, Social Security numbers, or ZIP codes. Even text strings of variable length like names are contained in records, making it a simple matter to search. Data may be human- or machine-generated as long as the data is created within an RDBMS structure. This format is eminently searchable both with human generated queries and via algorithms using type of data and field names, such as alphabetical or numeric, currency or date.
Common relational database applications with structured data include airline reservation systems, inventory control, sales transactions, and ATM activity. Structured Query Language (SQL) enables queries on this type of structured data within relational databases.
Some relational databases do store or point to unstructured data such as customer relationship management (CRM) applications. The integration can be awkward at best since memo fields do not loan themselves to traditional database queries. Still, most of the CRM data is structured.
What is Unstructured Data?
Unstructured data is essentially everything else. Unstructured data has internal structure but is not structured via pre-defined data models or schema. It may be textual or non-textual, and human- or machine-generated. It may also be stored within a non-relational database like NoSQL.
Typical human-generated unstructured data includes:
- Text files: Word processing, spreadsheets, presentations, email, logs.
- Email: Email has some internal structure thanks to its metadata, and we sometimes refer to it as semi-structured. However, its message field is unstructured and traditional analytics tools cannot parse it.
- Social Media: Data from Facebook, Twitter, LinkedIn.
- Website: YouTube, Instagram, photo sharing sites.
- Mobile data: Text messages, locations.
- Communications: Chat, IM, phone recordings, collaboration software.
- Media: MP3, digital photos, audio and video files.
- Business applications: MS Office documents, productivity applications.
Typical machine-generated unstructured data includes:
- Satellite imagery: Weather data, land forms, military movements.
- Scientific data: Oil and gas exploration, space exploration, seismic imagery, atmospheric data.
- Digital surveillance: Surveillance photos and video.
- Sensor data: Traffic, weather, oceanographic sensors.
The most inclusive Big Data analysis makes use of both structured and unstructured data.
Structured vs. Unstructured Data: What’s the Difference?
Besides the obvious difference between storing in a relational database and storing outside of one, the biggest difference is the ease of analyzing structured data vs. unstructured data. Mature analytics tools exist for structured data, but analytics tools for mining unstructured data are nascent and developing.
Users can run simple content searches across textual unstructured data. But its lack of orderly internal structure defeats the purpose of traditional data mining tools, and the enterprise gets little value from potentially valuable data sources like rich media, network or weblogs, customer interactions, and social media data. Even though unstructured data analytics tools are in the marketplace, no one vendor or toolset are clear winners. And many customers are reluctant to invest in analytics tools with uncertain development roadmaps.
On top of this, there is simply much more unstructured data than structured. Unstructured data makes up 80% and more of enterprise data, and is growing at the rate of 55% and 65% per year. And without the tools to analyze this massive data, organizations are leaving vast amounts of valuable data on the business intelligence table.
Structured data is traditionally easier for Big Data applications to digest, yet today's data analytics solutions are making great strides in this area.
How Semi-Structured Data Fits with Structured and Unstructured Data
Semi-structured data maintains internal tags and markings that identify separate data elements, which enables information grouping and hierarchies. Both documents and databases can be semi-structured. This type of data only represents about 5-10% of the structured/semi-structured/unstructured data pie, but has critical business usage cases.
Email is a very common example of a semi-structured data type. Although more advanced analysis tools are necessary for thread tracking, near-dedupe, and concept searching; email’s native metadata enables classification and keyword searching without any additional tools.
Email is a huge use case, but most semi-structured development centers on easing data transport issues. Sharing sensor data is a growing use case, as are Web-based data sharing and transport: electronic data interchange (EDI), many social media platforms, document markup languages, and NoSQL databases.
Examples of Semi-structured Data
- Markup language XML This is a semi-structured document language. XML is a set of document encoding rules that defines a human- and machine-readable format. (Although saying that XML is human-readable doesn’t pack a big punch: anyone trying to read an XML document has better things to do with their time.) Its value is that its tag-driven structure is highly flexible, and coders can adapt it to universalize data structure, storage, and transport on the Web.
- NoSQL Semi-structured data is also an important element of many NoSQL (“not only SQL”) databases. NoSQL databases differ from relational databases because they do not separate the organization (schema) from the data. This makes NoSQL a better choice to store information that does not easily fit into the record and table format, such as text with varying lengths. It also allows for easier data exchange between databases. Some newer NoSQL databases like MongoDB and Couchbase also incorporate semi-structured documents by natively storing them in the JSON format.
In big data environments, NoSQL does not require admins to separate operational and analytics databases into separate deployments. NoSQL is the operational database and hosts native analytics tools for business intelligence. In Hadoop environments, NoSQL databases ingest and manage incoming data and serve up analytic results.
These databases are common in big data infrastructure and real-time Web applications like LinkedIn. On LinkedIn, hundreds of millions of business users freely share job titles, locations, skills, and more; and LinkedIn captures the massive data in a semi-structured format. When job seeking users create a search, LinkedIn matches the query to its massive semi-structured data stores, cross-references data to hiring trends, and shares the resulting recommendations with job seekers. The same process operates with sales and marketing queries in premium LinkedIn services like Salesforce. Amazon also bases its reader recommendations on semi-structured databases.
Structured vs. Unstructured Data: Next Gen Tools are Game Changers
New tools are available to analyze unstructured data, particularly given specific use case parameters. Most of these tools are based on machine learning. Structured data analytics can use machine learning as well, but the massive volume and many different types of unstructured data requires it.
A few years ago, analysts using keywords and key phrases could search unstructured data and get a decent idea of what the data involved. eDiscovery was (and is) a prime example of this approach. However, unstructured data has grown so dramatically that users need to employ analytics that not only work at compute speeds, but also automatically learn from their activity and user decisions. Natural Language Processing (NLP), pattern sensing and classification, and text-mining algorithms are all common examples, as are document relevance analytics, sentiment analysis, and filter-driven Web harvesting. Unstructured data analytics with machine-learning intelligence allows organizations to:
- Analyze digital communications for compliance. Failed compliance can cost companies millions of dollars in fees, litigation, and lost business. Pattern recognition and email threading analysis software searches massive amounts of email and chat data for potential noncompliance. A recent example includes Volkswagen’s woes, who might have avoided a huge fines and reputational hits by using analytics to monitor communications for suspicious messages.
- Track high-volume customer conversations in social media. Text analytics and sentiment analysis lets analysts review positive and negative results of marketing campaigns, or even identify online threats. This level of analytics is far more sophisticated simple keyword search, which can only report basics like how often posters mentioned the company name during a new campaign. New analytics also include context: was the mention positive or negative? Were posters reacting to each other? What was the tone of reactions to executive announcements? The automotive industry for example is heavily involved in analyzing social media, since car buyers often turn to other posters to gauge their car buying experience. Analysts use a combination of text mining and sentiment analysis to track auto-related user posts on Twitter and Facebook.
- Gain new marketing intelligence. Machine-learning analytics tools quickly work on massive amounts of documents to analyze customer behavior. A major magazine publisher applied text mining to hundreds of thousands of articles, analyzing each separate publication by the popularity of major subtopics. Then they extended analytics across all their content properties to see which overall topics got the most attention by customer demographic. The analytics ran across hundreds of thousands of pieces of content across all publications, and cross-referenced hot topic results by segments. The result was a rich education on which topics were most interesting to distinct customers, and which marketing messages resonated most strongly with them.
In eDiscovery, data scientists use keywords to search unstructured data and get a reasonble idea of the data involved.
No matter what your business specifics are, today’s goal is to tap business value whether the data is structured or unstructured. Both types of data potentially hold a great deal of value, and newer tools can aggregate, query, analyze, and leverage all data types for deep business insight across the universe of corporate data.
Next steps: to fully understand the enterprise IT infrastructure that hosts today's structured and unstructured Big Data tools, read The Comprehensive Guide to Cloud Computing.