Editor’s note: Bill Inmon is a speaker for ODSC East 2025 this May 13th to 15th! Be sure to check out his talk, “Building AI’s Foundation: The Critical Role of Data Architecture,” there!

Modern technology – AI, ChatGPT, ML – is great. Modern technology has tremendous potential. Modern technology is built and serviced by vendors who make great promises about what the future can look like.

An Achilles Heel

But there is an Achilles heel to all modern technology. That Achilles Heel is that ALL modern technology depends on the assumption that there is a solid foundation of data on which to operate. Modern technology depends on a solid foundation of data, and modern technology depends on the premise that the foundation of data is available, complete, and accurate.

Yet there is no such foundation of data that exists in the corporation that is believable. The underlying foundation of data is a pig sty in most corporate environments. There are many reasons why the underlying foundation of data in the corporation is in such a terrible state of repair.

In order to understand how the underlying foundation of data got to be so disorganized and scattered, consider the evolution of the data found in the corporation.

Applications

In the beginning were applications. Applications were designed to suit the needs of a department or small collection of users. Soon, in the corporation there were a very large number of applications. And with this large number of applications arose a problem with the believability of data. In one application element ABC had a value of 20. In another application ABC had a value of 0. In another application ABC had a value of 675. Trying to make informed decisions on this motley and contradictory collection of values was impossible. Furthermore, trying to rectify what value was actually correct was even more difficult.

Certain organizations in the corporation – accounting, marketing, finance, and sales – needed to look across the corporation and create an enterprise-wide view of data. These organizations could not do it based on a collection of applications.

Data Warehouses

Into this morass came the data warehouse. The data warehouse made enterprise-wide data available to the corporation for the first time. With data warehouses, there was a single version of the truth.

The data warehouse was built exclusively on structured data. Structured data was a byproduct of the transaction processing systems that ran the corporation.

For a long time, data warehousing served the needs of the corporation quite well.

But over time, there started to appear data that was not structured in the corporation. In particular, there started to appear textual data and analog data. Textual data appeared in many forms – on the Internet, in emails, on a spreadsheet, in call centers, and so forth. Analog data appeared as a byproduct of machines processing some function and throwing off data as a byproduct. There were drones. There were metering devices. There were temperature gauges. Machines all over the corporation were generating data.

The confluence of the different kinds of data resulted in a really big mess of data in the corporation.

Different Properties

The issue with the different kinds of data – structured, textual, and analog – was that each of the different kinds of data had very different properties. The world was educated to expect structured data. And when a professional who had been schooled in structured data entered the worlds of text and analog, they were lost. The professionals who had lived with and understood structured data tried to use techniques and approaches they had learned over their career and those techniques and approaches simply did not work in the world of text and analog.

So what were some of the differences between these worlds?

• Structured data is a world of precision. Textual data is a stochastic world.
• Structured data employs explicit context. Textual data employs implied context.
• Structured data employs vertical contextualization. Textual data employs horizontal contextualization.
• Structured data employs identification of individual rows of data. Textual data employs classification of rows of data.
• And these are but a few of the fundamental differences between text and structured data.

Trying to use structured techniques and practices in a textual world is like asking an earthbound human to thrive and survive on Mars. The rules of living on Mars are far different than the rules of living on earth.

Analog Data

Analog data has yet another different set of rules and practices. In order to understand analog data, imagine yourself to be a security camera set up to do surveillance on a parking lot. Every 1/30th of a second you take a snapshot of the parking lot. Night and day. Seven days a week.

And what happens to the images that are created? During the night only a few cars are in the parking lot. During the day, the camera captures images of people parking their cars and arriving and leaving work. Yet the surveillance camera keeps taking pictures.

For all practical purposes, the images captured during these times are fairly useless.

Then one day – at 3:52 pm on a Wednesday in April – a car is broken into. The sixty seconds of images captured by the surveillance camera now become EXTREMELY valuable.

When you stand back and look at the analog data generated by the camera, 99.999% of the images are a waste of time and have little or no value. But .001% of the images of the surveillance camera are extremely valuable.

The phenomenon of the creation of a huge amount of data that has little or no value and the creation of a small amount of data that has extreme value is characteristic of analog data.

Trying to apply structured and textual techniques of data management to analog data is a fool’s game. Analog data requires its own unique treatment.

Data Architecture

In order to blend these very different types of data into a cohesive whole is what is required in order to produce the foundation of data that will support modern technology. That whole is called a data architecture.

When modern technology tries to deliver on its promises, it absolutely cannot succeed operating on a foundation of unreliable, unbelievable data.

Yet vendors and organizations ignore this fundamental and underlying fact. Vendors sell the sizzle. But vendors ignore the foundation that is required to make their technologies operate successfully.

Caveat emptor.

About the Author/ODSC East 2025 Speaker:

Bill Inmon is the father of data warehouses, named by Computerworld one of the ten most influential people in the history of computing, author of 72 books, books translated into nine languages, over 1,500,000 books sold, and inventor of textual disambiguation.