This chapter is from the book
This chapter is from the book
This chapter is from the book
1.2. Business Analytics Process
The complete business analytic process involves the three major component steps applied sequentially to a source of data (see Figure 1.1). The outcome of the business analytic process must relate to business and seek to improve business performance in some way.
)
Figure 1.1 Business analytic process
The logic of the BA process in Figure 1.1 is initially based on a question: What valuable or problem-solving information is locked up in the sources of data that an organization has available? At each of the three steps that make up the BA process, additional questions need to be answered, as shown in Figure 1.1. Answering all these questions requires mining the information out of the data via the three steps of analysis that comprise the BA process. The analogy of digging in a mine is appropriate for the BA process because finding new, unique, and valuable information that can lead to a successful strategy is just as good as finding gold in a mine. SAS, a major analytic corporation (www.sas.com), actually has a step in its BA process, Query Drilldown, which refers to the mining effort of questioning and finding answers to pull up useful information in the BA analysis. Many firms routinely undertake BA to solve specific problems, while other firms undertake BA to explore and discover new knowledge to guide organizational planning and decision-making to improve business performance.
The size of some data sources can be unmanageable, overly complex, and generally confusing. Sorting out data and trying to make sense of its informational value requires the application of descriptive analytics as a first step in the BA process. One might begin simply by sorting the data into groups using the four possible classifications presented in Table 1.4. Also, incorporating some of the data into spreadsheets like Excel and preparing cross tabulations and contingency tables are means of restricting the data into a more manageable data structure. Simple measures of central tendency and dispersion might be computed to try to capture possible opportunities for business improvement. Other descriptive analytic summarization methods, including charting, plotting, and graphing, can help decision makers visualize the data to better understand content opportunities.
Table 1.4 Types of Data Measurement Classification Scales
|
Type of Data Measurement Scale |
Description |
|
Categorical Data |
Data that is grouped by one or more characteristics. Categorical data usually involves cardinal numbers counted or expressed as percentages. Example 1: Product markets that can be characterized by categories of “high-end” products or “low-income” products, based on dollar sales. It is common to use this term to apply to data sets that contain items identified by categories as well as observations summarized in cross-tabulations or contingency tables. |
|
Ordinal Data |
Data that is ranked or ordered to show relational preference. Example 1: Football team rankings not based on points scored but on wins. Example 2: Ranking of business firms based on product quality. |
|
Interval Data |
Data that is arranged along a scale where each value is equally distant from others. It is ordinal data. Example 1: A temperature gauge. Example 2: A survey instrument using a Likert scale (that is, 1, 2, 3, 4, 5, 6, 7), where 1 to 2 is perceived as equidistant to the interval from 2 to 3, and so on. Note: In ordinal data, the ranking of firms might vary greatly from first place to second, but in interval data, they would have to be relationally proportional. |
|
Ratio Data |
Data expressed as a ratio on a continuous scale. Example 1: The ratio of firms with green manufacturing programs is twice that of firms without such a program. |
From Step 1 in the Descriptive Analytic analysis (see Figure 1.1), some patterns or variables of business behavior should be identified representing targets of business opportunities and possible (but not yet defined) future trend behavior. Additional effort (more mining) might be required, such as the generation of detailed statistical reports narrowly focused on the data related to targets of business opportunities to explain what is taking place in the data (what happened in the past). This is like a statistical search for predictive variables in data that may lead to patterns of behavior a firm might take advantage of if the patterns of behavior occur in the future. For example, a firm might find in its general sales information that during economic downtimes, certain products are sold to customers of a particular income level if certain advertising is undertaken. The sales, customers, and advertising variables may be in the form of any of the measurable scales for data in Table 1.4, but they have to meet the three conditions of BA previously mentioned: clear relevancy to business, an implementable resulting insight, and performance and value measurement capabilities.
To determine whether observed trends and behavior found in the relationships of the descriptive analysis of Step 1 actually exist or hold true and can be used to forecast or predict the future, more advanced analysis is undertaken in Step 2, Predictive Analytic analysis, of the BA process. There are many methods that can be used in this step of the BA process. A commonly used methodology is multiple regression. (See Appendix A, “Statistical Tools,” and Appendix E, “Forecasting,” for a discussion on multiple regression and ANOVA testing.) This methodology is ideal for establishing whether a statistical relationship exists between the predictive variables found in the descriptive analysis. The relationship might be to show that a dependent variable is predictively associated with business value or performance of some kind. For example, a firm might want to determine which of several promotion efforts (independent variables measured and represented in the model by dollars in TV ads, radio ads, personal selling, and/or magazine ads) is most efficient in generating customer sale dollars (the dependent variable and a measure of business performance). Care would have to be taken to ensure the multiple regression model was used in a valid and reliable way, which is why ANOVA and other statistical confirmatory analyses are used to support the model development. Exploring a database using advanced statistical procedures to verify and confirm the best predictive variables is an important part of this step in the BA process. This answers the questions of what is currently happening and why it happened between the variables in the model.
A single or multiple regression model can often forecast a trend line into the future. When regression is not practical, other forecasting methods (exponential smoothing, smoothing averages) can be applied as predictive analytics to develop needed forecasts of business trends. (See Appendix E.) The identification of future trends is the main output of Step 2 and the predictive analytics used to find them. This helps answer the question of what will happen.
If a firm knows where the future lies by forecasting trends as they would in Step 2 of the BA process, it can then take advantage of any possible opportunities predicted in that future state. In Step 3, Prescriptive Analytics analysis, operations research methodologies can be used to optimally allocate a firm’s limited resources to take best advantage of the opportunities it found in the predicted future trends. Limits on human, technology, and financial resources prevent any firm from going after all opportunities they may have available at any one time. Using prescriptive analytics allows the firm to allocate limited resources to optimally achieve objectives as fully as possible. For example, linear programming (a constrained optimization methodology) has been used to maximize the profit in the design of supply chains (Paksoy et al., 2013). (Note: Linear programming and other optimization methods are presented in Appendixes B, “Linear Programming,” C, “Duality and Sensitivity Analysis in Linear Programming,” and D, “Integer Programming.”) This third step in the BA process answers the question of how best to allocate and manage decision-making in the future.
In summary, the three major components of descriptive, predictive, and prescriptive analytics arranged as steps in the BA process can help a firm find opportunities in data, predict trends that forecast future opportunities, and aid in selecting a course of action that optimizes the firm’s allocation of resources to maximize value and performance. The BA process, along with various methodologies, will be detailed in Chapters 5 through 10.