Considering the massive volume of structured data streaming into databanks every day, it’s almost shocking to learn that it represents only 10 to 20% of all available data. The other 80 to 90% is unstructured data, the type of data that doesn’t fit neatly into tables, rows, or columns.
The disparity between the two types of data poses a significant barrier to organizations that want to analyze and extract all possible relevant information from their corporate information: today’s analytics programming can analyze structured data but not unstructured data. And without access to a process that converts unstructured information into structured, usable information, most companies can’t access the informational treasures hidden in those locked away files.
Everything is Information
Think for a bit about how much information flows into your consciousness on a daily basis. Today’s communities are awash with an immense variety of information presentation formats. Email accounts, radio stations, television shows, websites, streaming services, news feeds, newspapers (yes, they still exist), magazines, billboards, scrolling banners, and more are placed strategically in locations where they will be seen and consumed. Most of the information they share is irrelevant to daily business life, but some of it can be critical to corporate success and can even influence life or death situations. Unstructured data hidden in health reports, financial statistics, and even weather warnings can provide essential information that demands immediate attention. Problems, failures, and disasters are possible when the data are invisible.
Such negative consequences are even more likely when the programming can’t use available intelligence simply because it is inaccessible. While digital science has done a great job developing products and services that capture and store all this unstructured material, it’s not done as well at making it accessible and usable.
Today’s analytical programming requires a structured data format for it to perform its analytical magic. Data that isn’t structured for analytical purposes can’t be included in the analytical processing. And without access to all relevant information, any analysis will be skewed because its source material is limited. And the challenge of managing all this unmanageable information is growing, too. The volume of unstructured data is growing three times faster than the volume of structured information, so harnessing that data supply will only become more complex as its volume multiplies.
What is ‘Unstructured’ Data?
Unstructured data can’t be organized into a traditional, ‘relational’ format; it is composed of too many variables to be measured so precisely. The organization of a grocery store inventory is an example of how a relational structure lends itself to analysis:
- A grocery store provides a fairly standard collection of food categories – produce, meat, dairy, etc.
- Each of those categories can be grouped into similar types (vegetables, fruit, etc.), and then,
- those categories can be parsed even more finely (green produce, canned produce, boxed produce, etc.).
- The aggregate numbers in each category can then be filed into tables with horizontal and vertical axes.
The analysis program analyzes the table to determine how its various elements ‘relate’ to each other so that the store owner can track inventories of canned versus fresh vegetables, as an example.
Unstructured data, on the other hand, doesn’t come in a format that can be entered into a table. There are eight general categories of unstructured data, and within each are a myriad of sub-varieties:
- ‘Documents‘ are actually unstructured data. This category includes the many forms of ‘written’ documents available digitally, such as Word dox and text files.
- Email is a form of unstructured data. Each email contains the specific information of the sender and the recipient(s), the internal content, attachments, images, and other add-ons.
- Social media data comes in distinctly different formats from documents, and each social media program presents its own individual type of data format.
- ‘Machine data‘ – server, website, and application usage data (logs) generate trillions of data bits each day as users access their pages.
- Sensor data is also unstructured. They are generated by the 14+ billion Internet of Things (IoT) devices that sit silently and often hidden while collecting the information relevant to their purpose. This number is expected to top 27 billion by 2025.
- Video files generate an almost unmeasurable volume of unstructured data every second. Media sites such as YouTube, Facebook, and WhatsApp – all incorporate video – each service reports more than 2 billion individual users every day.
- Audio files, as stand-alone files or in conjunction with video or still images, generate another vast volume of information each day.
- Images are also an independent data format. There are an estimated 750 billion images saved in Internet files, 92% of which were taken with a smartphone. Another 300 million are uploaded every day.
Obviously, the challenges posed trying to capture and store all these various formats are staggering. The sheer volume of existing unstructured data makes gathering and keeping it very difficult. The problem is compounded because the data (usually) can’t be sorted or categorized in the same way grocery store inventory is organized. Typically, streaming unstructured data is captured ‘as is’ and stored in data warehouses designed solely to facilitate those functions.
Gathering the intelligence these files contain is even more difficult. Often, individual data records can have many uses, so sharing them with everyone who can use that information enhances their value. The number of cans of green beans sold, for example, reveals:
- the popularity of canned goods, which can influence the production volume of aluminum cans;
- the popularity of green beans, which can influence the volume of green beans produced, and
- the percentage of grocery shelf space which can/should be devoted to canned green beans.
Sharing all this data with every entity that needs it is an immense task that is often beyond the capabilities of most IT departments.
To learn how we helped a client by developing an innovative process to convert its large unstructured data lakes into usable, analyzable information, download our case study, “Finding Gold: Accessing Your Unstructured Data.”
Many companies turn to data management experts such as Datavail to tackle these complex information transformation projects. Datavail’s global panel of data transformation professionals is available 24/7 to ensure your organization gains control of all of its intelligence, even that hidden away in your unstructured data warehouses.
Check out these BI Publisher tips including functions & calculations so you can understand more about the production and support of BI Publisher reports.
Ultimately the goal of commentary in OBIEE is to have a system for persisting feedback, creating a call to action, and recognizing the prolific users.