Bringing It All Together (Data Mining on an Enterprise Level)

Up to now we have presented the fundamental building blocks to understanding the concept of data mining and addressed the prevailing applications within the corporate environment including both the “brick and mortar” style and e-commerce spectrums. The process does not stop here however. In order to implement mining on an enterprise basis, firms must overcome some potentially serious obstacles and address key issues. The more complex nature of data mining generally limits its use to a smaller population of individuals in a given firm, (although this is not always the case). Because of this, a common drawback to the process of effective Mining is the communication of value-added model results to corresponding users of this information. Just as there exists a gap between IT personnel, (those who know the technical side of systems) and the business user, (those who require IT systems to help solve their problems), there also exists a communication gap between the “data miners” and those who need to apply the resulting models to help solve their business problem. Other issues which must be considered before implementing an organization wide mining approach entails the development of total mining solutions instead of limiting applications to a few business problems. Decision makers must also avoid the trap of relying too heavily on mining results and must remember that these models are not crystal ball providers of perfect knowledge. Because of this, they must therefore monitor actual business performance against projected measures to maintain model effectiveness and accuracy.

What the Future Holds for Data Mining

Predicting the future is always a difficult task, of course that depends on how far into the future one attempts to delve. With regard to data mining, there’s no doubt the future should entail some interesting new applications that seek to enhance the process of discovering patterns and relationships existing between variables underpinning a given business application. This chapter seeks to enlighten the reader with regards to “what’s in the pipeline” for the coming years in the world of data mining. This topic can be broken down into two major components which include: 1) Innovations in statistics and algorithms that will provide new revelations to the world of mining. 2) Innovations in overall information technology that will augment the current functionality of mining methodology. This chapter will emphasize the second point mentioned above as the area for the greatest potential for mining enhancements over the next year or so.

Turning Your Brick and Mortar into a Click and Mortar

I recently read an article about words and terms that evolved exclusively from the American culture. This piece traced the history of American-born verbiage throughout the 1900s and into the year 2000. Not surprisingly, many of the words that appeared in later years were some of the most pervasive buzzwords and terms of our Web-wild culture: e-business, e-commerce, click-and-mortar, among others. It’s daunting, really; no wonder some retailers are confounded by what faces them as they ponder the move from offline to online. You’ve spent millions of dollars implementing a customer relationship management system to better understand your customers: their wants, their desires, their buying habits. You’ve used it to great success to build excellent offline customer relations. You’re now looking for the next big opportunity, the jump into something beyond what you’re currently doing. The truth is that those that do make the jump can be exposed to a tremendous amount of opportunity and increased revenue. Using the data that your customer relationship management (CRM) solution has collected about offline customers and “multichanneling” it, as it’s called, can lead to big profits. Traditional marketers’ eyes light up when they see forecasts predicting that $199 billion will have been spent by consumers online by 2005 (“Online Influencing Offline: The Multichannel Mandate,” Jupiter Communications, June, 2000). The challenge presented to marketers is how to successfully make the jump to promoting their goods in cyberspace. The thing to remember is that a lot of the same rules apply—but on the Web, you have a lot more options. Be aware that Web and offline marketing share some basic truisms. The ultimate goal is to identify your appropriate audiences and market to them accordingly. Get to know your audiences, present them information that is customized to their needs and interests, and deliver it to them in a non-intrusive manner. The Web makes it exceedingly easy to do this, especially with solutions like online profiling, which this chapter will cover in detail.

A Closer Look at Marketing/Advertising, Promotions and Pricing Policies Using Econometric Based Modeling

Two core business strategies throughout the realm of commerce, which take their root from traditional economic theory, involve the incorporation of marketing, advertising and pricing policies for corresponding products and services. The determination of optimal strategies for each of these concepts is crucial since they account for the success or failure for a particular product or products and potentially the well being of the organization. As you well know, even the best product or service that has been mismarketed or inappropriately priced has little chance to achieve success in the market place. This is illustrated by the recent success of Ford Motor Company and their implementation of smart pricing.

Data Mining Defined

Over the years, the term data mining has been connected to various types of analytical approaches. In fact, just a few years ago, let’s say prior to 1995, many individuals in the software industry and business users as well, often referred to OLAP as a main component of data mining technology. More recently however, this term has taken on a new meaning and one which will most likely prevail for years to come. As we mentioned in the previous chapter, data mining technology encompasses such methodologies as clustering, classification and segmentation, association, neural networks and regression as the main players in this space. Other analytical processes which are related to mining, as defined in this work, include such methodologies as Linear Programming, Monte Carlo analysis and Bayesian methodologies. In fact, depending on who you ask, these techniques may actually be considered part of the data mining spectrum since they are grounded in mathematical techniques applied to historical data. The focus of this work however, revolves around the former more core approaches. Regardless of the type of methodology, data mining has taken its roots from traditional analytical techniques. Enhancements in computer processing, (e.g., speed and processing power) has enabled a wider diffusion of more complex techniques to become more automated and user friendly and have evolved to the state of our current data mining.

An Introduction to Information Technology and Business Intelligence

The world of commerce has undergone a transformation since the early 1990s, which has increasingly included the utilization of information technologies by firms across industry sectors in order to achieve greater productivity and profitability. In other words, through use of such technologies as mainframes, PCs, telecommunications, state-of-the-art software applications and the Internet, corporations seek to utilize productive resources in a way that augment the efficiency with which they provide the most appropriate mix of goods and services to their ultimate consumer. This process has provided the backbone to the evolution of the information economy which has included increased investment in information technology (IT), the demand for IT labor and the initiation of such new paradigms as e-commerce.

Improving the Web Experience through Real-Time Analysis (A Market Basket Approach)

The previous chapter introduced the evolution of the information economy as it addressed the progress of commerce from “brick and mortar” to “click and mortar” corporate initiatives. The key to the success of this process lies in the management of data by transforming it into usable information and applying appropriate business strategy. This chapter provides a natural extension to this topic as it describes the process by which organizations can achieve success on the Internet through the use of data, technology and sound management tactics. E-commerce vendors face two important challenges: driving up purchases and maintaining customer loyalty. However, only 2.7 percent of browsers buy from any given Web site and only 15 percent of those buyers return to buy again (Forrester Research, Inc.). To succeed, e-marketers must find ways to keep visitors on their sites. They must make the visitors’ experience convenient, satisfying and personally relevant. Above all, they must entice Web visitors to come back for more.

Steps to Success for the Mining Process

The previous chapters have given you some background on the core components of corporate IT systems along with software technology that promotes “business intelligence” throughout an enterprise. This included a good foundation on the high end analytical portion of information systems, namely data mining technology. All this sounds fantastic, state-of-the-art software that helps increase the flow of value-added information which leads to a reduction of uncertainty in a given business environment. However, the bottom line to the productivity enhancing process from IT implementation really entails proper management and utilization of this technology. In other words, an organization can spend huge sums of dollars on the best systems available, but if they are not implemented properly, their value and dollars invested become useless. Data mining technology is no exception. In fact, because of the more complex nature of the technology (e.g., statistics and mathematic underpinnings), the potential for underutilization or improper utilization is probably greater than other types of analytical applications. The following chapter provides some helpful hints on how to manage the mining process as it illustrates some common pitfalls that exist in conducting a high-end analysis. Remember, today’s technology is good, but it doesn’t do all the work for you.

Essential Mining Approaches to Problem Solving

Forecasting and “what if” mining generally incorporates the application of regression and neural network methodologies. In certain cases, for more simple applications, univariate forecasting methods can be used. Forecasting procedures are more affiliated with time series data or historic data that extend back in time (e.g., monthly periods over several years). Other mining applications involve examining a section of data over a specified time period, (e.g., looking at a number of customers, employees or processes over a given time period, let’s say a six-month period). This approach is referred to as a cross-sectional analysis mentioned briefly in the last chapter. The following section will describe these mining approaches in a bit more detail to give you an idea of not only how to effectively implement them, but also when and in what situation you may need to apply them.