CHAPTER 2

Nature of Information

What do we mean when we use the term information? Are we thinking about its content and form, or how is it processed or transferred? Defining information depends on the perspectives of the provider and the potential user. What is useful information to someone in the marketing department may be meaningless to someone working in the maintenance department. Yet, they both require some form or type of information to perform their respective jobs. In a recent paper, Denning and Bell pointed out that information processing technology does not depend on the meaning of the information for the technology to work.¹ The implication is that both information of great value and that of little value are handled in the same way by the technology.

The meaning of the information depends on the observer. I would add that it also depends on the provider of that information. Denning and Bell describe the situation we are considering here well with their observation, “Moreover, the concept of information seems fuzzy and abstract to many people, making it hard for them to understand how our information systems really work.” Knowing what to measure and how to measure it in a useful way is a major challenge for many businesses. We will cover this important topic further in chapter 4 when we discuss how to acquire and manage the information needed by a business. For now, let’s talk about what is information and how it can be classified and characterized, particularly for the best cost-effective use by today’s enormously capable information technology.

In the context of this monograph, we need to recognize that for IT systems to process and transmit information it is necessary to convert that information to a format that computers and their peripheral equipment can use—that is, numerical values. This need for conversion is not only new and driven by the relatively recent development of electronic systems, but also a long-standing universal need to allow people to exchange information effectively. While we can compare two persons as to which one is taller when they are standing side by side, that determination is much more difficult when neither person is present, even more so if we may have not seen one or both of them in person. But, if we know their heights as measured in some standard unit such as inches, we can readily perform the comparison without their presence or ever having seen either of them.

In another example, if we want a color image that will appear to be the same no matter where it is printed or displayed in the world, we must be able to quantify colors in some way that can be used by different printers or displays to produce the same result. Moreover, that way should be easy to implement and preferably transparent to users.

Recognizing the need to convert between analog and digital information so that computers can process the information, move it from place to place, store it, and retrieve it for later use seems simple enough. But, and this is a big but, this sometimes is very difficult to do, particularly when dealing with complex data combinations such as images and books. Those combinations become much more difficult to define for search engines or to another person who has not seen the image or read the book.

So, what are the more important concepts we should be grasping at this point other than that data must be in numerical form for computers to handle it? First, if we can convert attributes and other descriptors into numerical forms, we can convert those forms back into attributes and other descriptors when we are done with processing, transferring, or storing them. Recognizing this allows us a wider range of choices and ways in how information can be used to improve business processes.

Classifications

To enable us to improve how we use information in our business processes, we need to increase our understanding of the nature of information. Let’s consider the following list of possible classifications that we could use to describe information:

To keep our discussion manageable, let’s confine it to the major classifications, and you can add descriptors later to improve the match to your particular situation.

Data Types

We begin with a discussion of data types. Things can be quantified using numbers or numerical values or assigned attributes that distinguish one item or concept from another or at least categorize them into groups or sets.

While number and numerical values are often treated as meaning the same thing, there is a subtle difference in how they can be used. For example, a number can be assigned to an object (serial number) or person (government ID or phone number) to identify it or him/her as being different from another object or person. Numbers can also be used to indicate rank, class, or category according to some agreed-upon rules among the users of such information. Hence, the processing operations possible with such numbers are matching records in different places for the same object or person, comparison, sorting, and classifying. Numerical values come from a continuum of values that can be processed mathematically to provide new information such as averages, variability, totals, and so forth. Where these distinctions become fuzzy is when part of an ID number or phone number is used to also indicate the geographic location of the person when the number was assigned. While this was useful information in the beginning, it is fairly useless in today’s mobile society. Another fuzzy distinction is when numbers are assigned to a survey scale where 10 represents excellent and 0 represents terrible or a medical pain assessment by a patient where 10 represents excruciating pain and 0 represents n0 pain is felt. In these instances, there is a temptation to calculate the average level of service experienced by patrons or pain level felt by patients. However, one has to question the usefulness of knowing these values since it is unclear as to how much the average represents the average of assigned values or the variability in the assessment process assigning those values. Furthermore, when dealing with individual customers or patients, one needs to recognize that each of them expects to be considered according to the level of their personal experience, not the average experience of others.

In chapter 1, we discussed the development of information theory² where the focus was on the elements required to transfer information from one place to another. During that same period, another often-referenced paper on the theory of measurement was published in 1946 by Stevens, where he defined four areas of measurement. These areas are summarized briefly in Table 2.1 along with some more areas such as log-interval, cyclic, and multidimensional added later by Stevens and others.³ It should be noted here that the rapid advance of information technology will likely create the need for additional areas of measurement. Understanding the differences between each area is important because it affects how that data can be evaluated.

Table 2.1. Areas of Measurement

To emphasize this importance, I often asked my sophomore students to complete a simple assignment where they were provided several sets of data and were asked to analyze the data in whatever way the students felt appropriate and then submit a summary report in a manner that the students would expect their future bosses to want. Many of the students would jump right in and begin calculating averages, standard deviations, median, minmax values, and ranges without much regard to what the numbers in the data sets really represented. Only a few of them would think to graph the data, do frequency distributions, consider variances, or sort the data. Two of the data sets that they were asked to compare had identical averages, medians, standard deviations, and ranges, but distinctly different frequency distributions—one normal and the other bimodal. Typically, less than half of the students caught the difference.

As we discussed earlier, information must be expressed in a form that computers can compare (necessary for logic functions), add, subtract, multiply, divide, and, perhaps, most important, store. In a later work, Stevens summed up on page 25 how to meet this need with his definition “measurement is the assignment of numerals to events or objects according to rule.”⁴ Other higher-level logical and mathematical computations are made possible using combinations of these basic operations in algorithms that perform logic operations or approximate various algebraic and trigonometric functions. Some of these functional algorithms are common power series expansions, resulting in defined sequences of additions, subtractions, multiplications, and divisions used to determine the value of a function using a single input variable.⁵ The challenge, of course, is converting information such as the display of temperature in a glass thermometer into digital bits (ones and zeros) that computers can manipulate. The rapid development of analog-to-digital converters and a wide array of sensor technologies since the implementation of computer-controlled manufacturing systems in the 1970s enable businesses and industries to meet such measurement challenges.

Attributes such as big, small, tall, short, red, green, good, bad, poor, grades, addresses, and so forth, are simpler forms of unstructured data because they can be interpreted variously by each observer or user of the data. Your perception of a tasty dessert is quite likely to be different from my perception of the same dessert, particularly if it is of a flavor that I happen to dislike. The color of an object as perceived by a human observer varies from individual to individual and agreeing on whether or not something appears to be a particular shade of blue is very difficult if there is no standard paint chip reference or ceramic tile of that color to compare the sample with. This becomes more difficult when communicating or storing information regarding color and more so when reproducing a color from that information. There are different formats depending on how the color is presented. Computer displays, on the one hand, often use an additive approach such as RGB (primary colors of red, blue, and green) where black is no presence of those colors and white is the full addition of all of the colors. Printers, on the other hand, have to cope with a white background rather than a black background when no color components are present. Hence, they use a subtractive approach such as CMYK (cyan, magenta, and yellow colors aligned with a black key). A black shade could be obtained by just printing all of the CMY components in the same place, but the black would not be as pure as just printing a pure black pattern (the key).⁶ The ultimate challenge, of course, is being able to have a user’s print output look the same as the display on the user’s computer!

Like visual attributes, other attributes can be assigned numbers or alphanumeric characters so they can be processed, transferred, and stored electronically. Careful attention to these assignments is required because they are where many businesses encounter most of their database errors and inconsistencies. Although there are regulated standard units of measure for dimensions, weight, currency value, frequency, time, energy, and color, it is left to each individual business how they want to quantify or otherwise enter attributes into their database. To illustrate some of the difficulties, just consider how you would enter defect data for returned products or how to process customer complaints entered on your web site. Would a coworker enter the same information in the same way? If not, could you still analyze the information effectively? We will discuss this important issue more in chapters 4 and 6.

Finally, let’s discuss coded data. While some may feel that coded information should be treated as a data format or form, I consider code to be a higher form of expressing information where additional translation is required to access the true meaning of the information. When information is encoded into bits and bytes of data (sequences of ones and zeros) the basic translation process follows a set of rules that are either widely known or can be readily discerned. Therefore, it is much easier for an unintended person to intercept, read, and even alter the information during transferal or access from storage.

Using Shannon’s information communication process as illustrated in Figure 1.1, unpredictable disruptions or interferences in the transmission path are called noise. Some noises such as weather conditions or power interruptions over the transmission path can cause bit errors in the data, affecting its accuracy. Unauthorized access and alteration can also be considered to be a form of noise. It is also important to recognize the presence of noise in the conversion processes at both the sending and the receiving stages of Shannon’s model. I consider the sending conversion to be the most critical since any error in translation here will be propagated through the remaining stages even if they perform their respective functions without error.

Two types of coding are used to combat these different types of noise. One is the practice of appending some bits of data called parity bits to each transmitted packet of data to help the receiver identify which packets were received correctly and which packets were not. This form of coding is used when reading barcodes to ensure the barcode is read correctly. A simple version of this is counting the number of ones that are in the data packet before it is sent and adding another one at the end of the packet if the sum is odd or a zero if the sum is even. The result is that each coded packet has an even number of ones and the original data can be recovered by discarding the last bit. The receiver then counts the number of ones in each packet and asks the receiver to resend any packet that has an odd number of ones. This simple scheme detects single bit errors very well, but is insensitive to an even number of bit errors in a packet. It also slows down the overall transmission rate when the noise level is high because of the many resend requests. More exotic error detection schemes such as cyclic redundancy checks (CRCs) have been developed that not only detect a wide variety of bit errors, but also provide the receiver with enough information to correct those bit errors without having to request a retransmission of the data. Modern computer disk drives rely heavily on this type of coding to ensure reading and writing accuracy. However, because these error-correcting codes are widely documented they are not effective at preventing intruders from reading or altering the data.

This brings us to the next level of coding designed to prevent unauthorized access to information. More commonly referred to as encryption, the rules for coding and decoding are either known only to the sender and receiver or are based on a process using elaborate calculations that can be readily solved if one knows the key. We will discuss encryption in more detail in the cyber security concerns portion of chapter 6. For the moment it is wise to recognize that given enough time any computer security setup or data encryption can be broken if an individual or organization wants to do it. The challenge, therefore, is to make the time required long enough so that it becomes impractical to make the attempt.

Format

As the amount of data acquired increased, better ways of presenting it were explored. Computer routines replaced the tedious manual computations required to translate columns and rows of numbers into displays of graphs and charts for faster interpretation and analysis. A useful reference for other ideas on how quantitative information can be interpreted and displayed is The Visual Display of Quantitative Information, a book by Edward Tufte. It was originally published in 1983 just as personal computers were beginning to make their presence felt in the workplace. Since then, Tufte has updated the original book⁷ and written three other books on the subject.⁸ All of them are useful reading, particularly the latest one in 2006 where he critiques and provides useful suggestions for improving many of the ways businesses present information. As an example, one of his ideas, “sparklines,” was first introduced by him in one of his forums on May 27, 2004. While small word-sized graphics have been used for more than several centuries in printed material and even in hieroglyphics to convey information, Tufte expanded their use to display extensive sets of data in a compact way for spreadsheet formats. He describes sparklines as “datawords: data-intense, design-simple, word-sized graphics.” Their effectiveness in communicating information has resulted in their use in a number of publications and software data processing programs. For example, the 2010 version of Excel provides a function labeled “sparklines” in its Insert menu where the data in a range of cells can be graphed easily within a single adjacent cell to show trends in the data. Figure 2.1 shows two sparkline results on the right for the same range of data listed in the seven cells to the left to demonstrate how numerical information can be communicated as an image.

Figure 2.1. Sparkline example from Excel.

Another way to show trend data would be to use an Excel function to highlight each cell with a color signifying whether the trend from the previous period is up (green), flat (yellow), or down (red). This is referred to in Excel’s references as conditional formatting or data visualization.

Other ways to indicate changes in data are different tones to indicate different operations or status, different vibrations to indicate the nature of an incoming text or voice message on a cell phone, flashing lights of different colors to indicate the presence of different situations, music choices to support different moods, or even odors to indicate danger like the additive used in natural gas to indicate a leak.

Aggregation is in my mind one of the most important formats for business. Although one of the Big Data rationales is to include more and more data for analysis, that increased level of details can obscure more obvious trends and imply a level of precision and accuracy inappropriate for estimating, forecasting, scheduling, or purchasing processes. Also, it requires more storage capacity, computing power, and increased data entry activity—all adding to operating costs to offset any perceived benefit. Aggregation is basically combining unit-sized data to larger, more manageable collections of that data. In its most basic form, it is rounding off data with decimal values to whole numbers when a fractional component is not likely in reality, that is, a fraction of a person, TV set, or keg of beer, for instance. A more practical application is combining data such as demand into quantities appropriate for handling, inventory, or purchasing. Examples are cases of wine, boxes of donuts, forklift pallet loads, barrels of oil, and tons or hundred weights of raw materials. Other less obvious aggregations are the total number of trucks, appliances, fast-food meals, and so forth, produced when the majority of their components and labor content are essentially the same. That is, tracking the number of black trucks sold does not affect how much steel or labor is required, only the amount of black coloring required for the paint. This is particularly true in today’s supply chain environment where the majority of the manufacturing either uses all standardized parts or can assemble-to-order using a few special components from outside suppliers. In the latter case, those specialized items are usually standardized items for the suppliers and only need to be tracked by the buyers in whatever order sizes the supplier chooses to deal in.

All of these format possibilities can be used at either the sending or the receiving end of the information communication process described by Figure 1.1. To transfer information from one location to another requires some additional consideration of the form or media used for the transfer because some forms are not suitable for some formats.

Form/Media

The physical manifestation of information greatly affects how it can be used, transferred, stored, and retrieved. Long, long, ago in the Mesopotamian region of Asia, the Sumerians recorded information by making marks in a wet clay tablet, which when dry became not only a storage medium for that information, but also provided a means for transmitting that information from place to place. While considered to be a primitive method today, it should be recognized that a number of these clay tablets have survived for more than five thousand years and can still be read today by a few scholars, an achievement that today’s storage technologies will find hard to beat when considering that the lifetime of their media is immensely shorter and the technology required to read it becomes rapidly obsolete in less than a century—78-rpm records, wire recorders, tintype photographs, 8-track cassette tapes, 5¼-inch floppy disks, Betamax video tapes, and 8mm movie films are some of the many examples.

Dematerialized information in its purest definition is a form of energy—light, heat, infrared radiation, radio waves, electric current, sound, scent, and so forth. In that form, it is easily transmitted, but very difficult, if not currently impossible to store for any length of time and not easily observed. Therefore, for most practical purposes, its definition must include the use of some physical systems that can process, display, and store digitized information electronically. For businesses using cloud computing resources that physicality is for the most part located somewhere else; so it could be said they have come closest to truly dematerializing their data.

Transferal Method/Medium

Information in material forms such as clay tablets, knotted string, paper, films, photographs, blueprints, paintings, tape recordings, CDs, DVDs, and flash drives requires physical transport by courier, horse, automobile, truck, train, carrier pigeon, plane, or rocket unless the information can be converted into some form of energy pattern that can be transmitted and received. Hence, the form of the information determines how it is communicated and processed, not the essence of the information itself. Furthermore, the choice of which form to use for the information is the major factor affecting how fast and accurately the information can be transferred.

The dematerialization of most of the information used by business and individuals has measuredly affected the methods used to transfer information in the past decade. Up until the mid-1990s, most of that information was printed on, transferred, and stored in the form of paper documents. In effect, we used the same transferal methods as we used to transfer physical products and the raw materials and supplies used to make them. Facilities were distributed geographically to shorten delivery time, both of products to customers, and to be closer to suppliers. A disadvantage of this strategy was that it was more difficult to share timely information among the various locations of a business. Overnight express mail services came into being for more urgent transferals of information paperwork such as contracts, orders, and invoices. The beginnings of text-only e-mail in the form of internal company systems for large enterprise companies that could afford them began to spread in the 1980s.⁹ This enabled higher-level executives to exchange rough drafts of contracts and orders more quickly for review and initial approval before preparing final paper versions for mailing and signatures. Other, less urgent business information and information for potential customers (advertisements) were sent via the local postal service. For those of you who have begun your careers after the mid-1990s, it is likely hard to imagine how much time was required for decisions to be made when the time to communicate information was based on moving paper around.¹⁰ Today’s ability for businesses to use accounting services anywhere in the world and to have immediate access to operational data at any location would not be possible without the widespread dematerialization of information and the concomitant ability to transmit it electronically using microwave and satellite communication systems.

Some method of conversion has always been necessary to transfer information by some physical conductor or a nonmaterial means from one place to another. Before the recent electronic advances in information technology, this conversion was a tedious process and the dematerialized form used for transmission was difficult to store for later retrieval. Morse code in its simplest form used a person at the transmitting end of an electrical conductor or radio communication system to translate letters and numbers into sequences of dots and dashes using a telegraph key (often expressed as dits and dahs to better represent the difference in duration of these two basic elements of the code). Another person at the receiving end translated the resulting sound sequences created by a relay device activated by the electrical pulses into the letters or numbers they represented. The speed of transmission was based on skill, with a good person at each end being able to transfer information at a rate greater than 100 characters per minute.¹¹

Today this process is accomplished at blinding speed using computers to translate the information into electrical signals representing the ones and zeros of much more complex digitized data. The technology that enables this to happen so that we can readily communicate all types of data with any part of the globe in a time limited only by the speed of light is far too complex to explain fully here. What does concern business professionals is its availability to their business. This is unlikely to be of concern to companies located in larger metropolitan areas, but can be a significant limitation to small businesses located in less densely populated areas where broadband and cell phone services are still sparsely provided. We will discuss these aspects in more detail in chapter 5 with other managerial considerations.

Purpose

The intended use of any information determines the best combination of type, format, form, and transferal method to use. Some information is more static in nature, has less time urgency, and may need to have some physical format for convenient use. Examples are a set of instructions, a user manual, safety warnings, restaurant menus, reference data, and archival records. An exception here is those records such as personal medical files and dangerous watch lists that need to be accessible quickly on a global basis.

Operational data used to control service and manufacturing processes must be timely, of sufficient accuracy and precision, and in a format and form appropriate for the application. Increasingly, this type of data must be online in digital formats both for automated manufacturing systems and for rapid service responses. Measurements of inputs, intermediate steps, and outputs of these processes are also necessary for direct control and for providing information to monitor process quality and improve process performance.

Customer information is becoming increasingly important to businesses. Initially its purpose was confined to payment, shipping, and warranty information that was easily handled in a structured manner. This changed when retailer, advertising, and marketing professionals realized they could combine increased IT computation capabilities with the dematerialized information stored in customer databases to mine such data for customer preferences, targeted advertising, and demand forecasts. This led to a number of new issues to address such as increased security risk, database duplications, errors, and storage capacity. These issues and others are covered in more detail later in chapters 3 through 6.

Information Characteristics

Each of the aforementioned information classifications can have various characteristics such as:

Take a moment and think about what you would add to these descriptors because the wide range of possibilities makes it easy to forget something. IBM authors use several V-terms (volume, velocity, and variety)¹² to describe data in their book on Big Data (Zikopoulos et al., 2012), a topic we will discuss in more detail in chapter 5. One of their sales representatives added veracity to these descriptors in a recent presentation about Big Data to IEEE members in my local section. I would add, with a smile? and keeping to the V-term theme, value, variability, versatility, and viability to the list.

Variability

At any given instant, a piece of information about an item or situation can be very accurate and precise, but an instant later can be another value equally accurate and precise if the item or situation is changing in some way. We will ignore the variations introduced by measurement errors and other information acquisition methods for the moment, but will discuss those concerns later. How much the information varies over time is described as deterministic if the variation is small enough to not affect the operation or decision using that information. If not, the information variability is considered to be stochastic, a collection of possibilities that usually have some boundary conditions but within those boundaries are unpredictable (occur at random). Occasionally we are lucky and able to observe that these possibilities occur according to some probability distribution. Such knowledge helps businesses make better decisions or develop more effective methods for handling such situations.¹³ A common example is the management of businesses where the service or production process requires the use of waiting lines or WIP queues because the customer arrivals or processing times are stochastic.

Accuracy and Precision

When we say accurate, we mean how closely the value represents the true value. Precision is how finely we define that value. For example, 3.14159 is an accurate value for pi; 3.14159265358979323846 is a more precise accurate value of pi. A properly exposed one-megapixel image of a person’s face can be an accurate representation of that person (barring the use of makeup, disguises, and so forth), but a 12-megapixel image of that person is a more exact precise representation. These distinctions are important because many people incorrectly use the terms precision and accuracy interchangeably. Thus, one can have a highly detailed image that is not an accurate representation of a person if it does not include a major portion of that person’s face.

It is important at this point to correct any misimpression that accuracy and precision are independent concepts when we consider processing numerical information in computer applications, particularly when more complex mathematical operations than addition and subtraction are required. If two groups within a company start with the same accurate information, but use different precisions because one group prefers to round off numbers to two decimal places while the other prefers to use six decimal places for all of its data, the outcomes for the same decision analysis by both groups will differ to a degree commensurate with the complexity of the analysis.¹⁴ As shown in Figure 2.2, this can occur in even a simple analysis such as calculating the area and circumference of a circle with a known radius. Here the value of pi is expressed according to each group’s precision. The effect of expressing pi to 15 decimal places is also shown, because each group may use Excel’s pi function in their calculations for convenience and are unlikely to recognize that in doing so they are mixing up the levels of precision used.

Figure 2.2. Precision example using Excel.

Some of you may be thinking that the differences in the results are pretty small, so what’s the big deal? Among other effects, there are two important things to consider. One, you can observe that the differences are more obvious for the more complex calculation where the radius is squared to get the area. In computations where there are a large number of complex operations, such differences can be significant. Second, we need to recognize that we are dealing with computer systems and databases here where they are operating at their own levels of precision and when they compare results with the results from another organization they will report errors unless the software developers allow some level of difference to account for precision errors. As we discuss in chapter 6, this requires some managerial attention to setting standards for data entry across different functions in a company. This is not an easy task since businesses often must use information that is neither accurate nor precise, as they would like, usually because there is insufficient time before a decision must be made to obtain more accurate and precise information. In such cases, managers must work with estimates or approximations or just make the best guess based on prior experience. With luck, they may have some tolerances provided with these uncertain values to help narrow in on the best decision or operational setting.

Structure

Information is either structured, has a defined range of values or choices of values, or unstructured, where the range of values can be very difficult to define, those values can vary widely and unpredictably, and some of the values can be interpreted differently by different users.

Another definition for structured information is data that can be readily identifiable. The most common examples are information that can be used in spreadsheets, drop-down menu choices, and relational databases. Both numerical and attribute data types can be used, but the attributes must be either in a form suitable for sorting or comparing such as addresses and names or must be convertible to numerical values for a computer to process them into categories or distributions.

Most people describe information as either numbers (data) or words or both. Although they watch and listen to the news on television or the Internet, they usually do not think of images and audio as being the most predominant forms of information today for the majority of the global population. Analyzing, storing, and retrieving such unstructured information is much more difficult because of the wide variation in its content. Some progress has been made with image and audio files by assigning descriptive tags to provide some structure, but since the number of tags must necessarily be limited, most images are not described well. Again, some managerial consideration is required for a given business to be able to handle its image or audio data effectively.

Longevity

How long information must remain available or be retained is a major business concern and often a major operating expense. Some information is ephemeral, needed only for the moment such as asking a customer in a coffee shop if they want room for cream.¹⁵ Other information may need to be stored for a considerable length of time to satisfy regulatory requirements or to provide a reference base for forecasting demand, long-term contract and insurance records, reviewing product failure causes, and so forth. These varying storage needs are an important part of information management, both to provide a business with the data it requires for successful operation and to support customer needs. More recently, longevity concerns have expanded to include electronic correspondence, particularly e-mail. In addition to compliance with changing governmental requirements regarding e-mail record retention, businesses also need to establish internal policies regarding electronic correspondence.

Security

Lastly and increasingly important, the security of information must be considered. Some information such as user manuals can be publicly shared without risk to the business, some information needs to be guarded to prevent a competitor from gaining useful knowledge regarding your business, and some information regarding critical assets such as customer ID data, financial accounts, and proprietary secrets must be secured. One negative effect of dematerializing information is an increased risk of unauthorized access and alteration of that information. Electronic forms of information are more easily transmitted, copied, and can be widely available. If connected to the Internet, such information is exposed to a worldwide level of attack.

Therefore, it is important for a business to classify what information can be publicly shared without risk to the business and to implement safeguards and access processes appropriate for the level of risk of unauthorized access or tampering of information that cannot be shared openly. We will discuss some of these strategies and policy considerations later in chapter 6.