Hello, I’m Matthew Schieber and I will be presenting the topics covered in chapter 4 about how organizations make decisions.
This chapter looks at the analytics processes and how analytics is used to help organizations make decisions. The chapter focuses mainly on decision support systems and artificial intelligence.
Every decision a person makes must first go through an analysis process. This decision making process takes place through four phases. The first phase, called intelligence, is recognizing an opportunity or a problem which needs resolving. In the second stage called design you consider the different solutions that could help to resolve the need. The third phase is where you make a choice and pick one of the possible solutions identified in the design phase. The last phase, called implementation, is where you actually carry out the choice you made in the third phase. For most people this process is done in a split second but this process can take days, weeks, or even longer before a final decision is made. The decision process is not linear, meaning the four phases do not have to happen in sequence. The phases can be repeated or even skipped.
Lets use buying a car as an example. For most people once the need to buy a car is determined will spend weeks in the design phase shopping for cars comparing deals, brands, dealerships, etc. One person might make the choice to keep their old car and do nothing. Another person might choose to buy the new Ford Fusion only to change their mind a few days later because Toyota offered an additional $1000 rebate causing the person to go back to the design phase. In the end whether a person takes action or no action a decision is made.
The types of decisions people make fall into one of four types of decisions. One type is a structured decision which are decisions that can be programed so you always get the same response. These decisions can be programmed into computers and are easily automated such as accounting processes or the retrieval of a forgotten password. On the other side of the coin is a nonstructured decision. These decisions do not have a best answer. These decisions often require a lot of intuition to determine what could be the best outcome. An example would be deciding what new product to launch or to develop a new marketing campaign for an existing product.
The other two types of decisions are based on how frequently a particular decision is made. Recurring decisions happen on a repetitive basis while nonrecurring decision are made infrequently. A recurring decision could be determining how many sales or promotions to offer in a year. A nonrecurring decision could be determining where to build a store.
Most decisions fall somewhere in the middle area as very few are purely at the extreme ends. Ideally most people would like to make decisions that are structured and recurring in nature. These are the easiest to make and the type of decisions organizations try to automate. The hardest and yet most important decisions to make are ones which are unstructured and nonrecurring. Determining where to build a store or even a house would fall into this area. These decisions have historically been all but impossible to automate but as technology continues to advance in the area of artificial intelligence, discussed more later on, even these types of decisions could one day become automated.
Ideally everyone would like to make a perfect decision based on complete perfect information. The reality is this is just not possible. Constraints such as time, money, and resources force people to make less than perfect decisions. Instead people make satisficing decisions in which a choice is made that fills a need but might not be the best possibility. Again lets use buying a car as an example. It is physically impossible to check every dealership in the whole country to try to determine which would give the best deal on the same type of car let alone different models or different brands. Instead people satisfice usually just checking dealers within 100 miles from home and buy a car which fills a need but might not be the best fit for that need.
Since businesses have to make thousands of decisions every day in a timely manner organizations have invested heavily into developing a decision support system (DSS). A DSS is a system that models information for users to assist in the decision making process. A DSS gives managers access to large amounts of data applying sophisticated analysis techniques to find specific information reducing the amount of time needed to make a decision. Without a DSS managers would have to sift through data manually.
DSS systems come in all shapes and sizes but most are typically made up of three main components. The first is a user interface which just allows a person to interact with the system via a GUI interface. The second component involves data management which stores and maintains information. The last component is model management which uses statistical and analytical models, such as goal-seek or what-if models, to present information in a way that can be understood by the user.
A geographic information system (GIS) is a very specialized DSS system that analyzes latitude and longitude information and shows it in map form. Some cell phone providers equip their phones with GPS chips that enable users to be located to within a geographical location about the size of a tennis court. Automobiles have GPSs linked to maps that display in a screen on the dashboard driving directions and exact location of the vehicle. GM offers the OnStar system, which sends a continuous stream of information to the OnStar center about the car’s exact location. The OnStar Vehicle Diagnostics of GM automatically performs hundreds of diagnostic checks on four key operating systems - the engine/transmission, antilock brakes, air bags, and OnStar systems. The vehicle is programmed to send the results via email to the owner each month. The unique email report also provides maintenance reminders based on the current odometer reading, remaining engine oil life, and other relevant ownership information.
For a DSS system to work it has to be connected to a database or data warehouse. But being connected to a database does not mean a person can find any data inside the database and use it. Just like how most people would not know how to access the almost limitless data on the Internet if they did not have a web browser such as Internet Explorer or Google Chrome. To gain access to the data inside the database DSS systems use data mining techniques to find specific data. No matter what technique is used data mining finds information in one of five different ways.
One way is by association in which two items are linked together. An example would be a person who has a drivers license also owns or drives a car. Another way is by clustering where things are grouped together without any rational reasoning. I recently read a news headline stating that people who eat at least three apples a day are twice as happy as people who don’t. A third way is by classification (prediction) where you look at historical data to help predict the future. Another way is by regression to find relationships between sets of data. The last is by summarization in which you perform simple math calculations such as sums and averages.
Data mining sometimes use specialized analytics to dig through data to find answers. One specialized type is called predictive analytics which is used to try to look into the future. In order to work properly predictive analytics needs two things. The first is a goal which is a question a person wants answered. The other is an indicator. Indicators give specific attributes to the goal. To work properly many indicators are used at the same time.
Another specialized form of analytics is text analytics. Text analytics uses some of the latest technology available to convert handwritten information into useful meaningful information a computer can read and analyze. Text analytics is far more complicated than predictive analytics because not only does it use the same processes as predictive analytics but it must also use linguistic technologies that look at how frequently words are used, meanings, emotions, and many other parts which can be very subjective.
Another area helping businesses make better decisions is the use of artificial intelligence (AI). As a field of study AI has exploded in the last decade as technology advances. The book talks about four main types of AI: expert systems, neural networks, genetic algorithms, and agent based technologies. Each of which will be looked at in more detail for the rest of the presentation.
The first and probably the biggest AI system businesses are looking at is expert systems. Expert systems help managers to use sound reasoning and logic to make decisions. Expert systems are best used for determining what is wrong when something happens and what to do to fix the problem. Expert systems help to reduce errors, reduce costs, improve customer service and to automate some decision making processes. Most expert systems contain information from many human experts and can therefore perform a better analysis than any single human. These types of systems are extremely important to businesses right now because of the aging work force and the retirement of the baby boomer generation. When these workers retire all of the knowledge gained during their 30 plus years with the organization walk out the door with them. Expert systems are a way for organizations to capture some of that knowledge so newer works do not have to spend as much time and effort in relearning the same knowledge.
Another type of AI system is a neural network. Neural networks are great at finding and differentiating patterns. Neural networks are used extensively in handwriting and speech recognition software. Some of the advantages of neural networks is the ability to self learn over time, the ability to analyze huge volumes of information that may be incomplete, and the ability to analyze nonlinear relationships. An important component is whats called fuzzy logic. Fuzzy logic allows the neural network to understand and assign values to subjective information. Just like fish stories terms such as large or gigantic vary from person to person.
Genetic algorithms are systems that find thousands or millions of possible solutions to a problem determining which solution is the best in a survival of the fittest process. It finds what is the most optimistic solution based on given set of constraints.
The last type of AI are agent-based technologies. Sometimes known as software agents these systems act on behalf of another person or program. Agent-based technologies are similar to how celebrities use agents to help maintain their profession. There are five main types of agents used today. Autonomous agents adapt with the environment as a means to complete assigned tasks. Distributed agents work on several computer systems at the same time. Mobile agents are able to move from one system to another.
Intelligent agents use reasoning and learning in order to perform tasks. Intelligent agents are used for environmental scanning and competitive intelligence. An intelligent agent can learn the types of competitor information users want to track, continuously scan the Web for it, and alert users when a significant event occurs. Intelligent agents also use data-mining agents to continuously scan a data warehouse to find information.
AI researchers are now modeling complex organizations as a whole with the help of multi-agent systems where a group of intelligent agents have the ability to work independently and to interact with each other when needed. Multi-agent modeling is a way of simulating human organizations using multiple agent-based technologies, each of which follows a set of simple rules and can adapt to changing conditions using biomimicry. These systems have become so powerful they can now perform swarm intelligence. Swarm intelligence is the ability to look at the collective behavior of all systems and develop solutions which work on a global scale. Large corporations use these systems to become leaner and more efficient by eliminating waste and to operate as one organization rather than several organizations using the same company name.