CHAPTER 4

Descriptive Analytics—Overview, Applications, and a Case

Overview: Descriptive Analytics

Descriptive analytics tools are used to understand the occurrence of certain business phenomenon or outcomes and explaining these outcomes through graphical, quantitative, and numerical analyses. Through the visual and simple analysis, descriptive analytics explores the current performance of the business and the possible reasons for the occurrence of certain phenomenon. Many of the hidden patterns and features not apparent through mere examination of data can be exposed through graphical and numerical analyses. Descriptive analytics uses simple tools to uncover many of the problems quickly and easily. The results enable us question many of the outcomes so that corrective actions can be taken.

Successful use and implementation of descriptive analytics require the understanding of data, types and sources of data, data preparation for analysis (data cleansing, transformation, and modeling), difference between unstructured and structured data, and data quality. Graphical/visual representation of data and graphical techniques using computer are basic requirements of descriptive analytics. These concepts related to data, data types, and the graphical and visual techniques are explained in detail in this chapter. The visual techniques of descriptive analytics tools include the commonly used graphs and charts along with some newly developed graphical tools such as bullet graphs, tree maps, and data dashboards. Dashboards are now becoming very popular with big data. They are used to display the multiple views of business data graphically.

The other aspect of descriptive analytics is an understanding of simple numerical methods, including the measures of central tendency, measures of position, measures of variation, and measures of shape, and how different measures and statistics are used to draw conclusions and make decision from the data. Some other topics of interest are the understanding of empirical rule and the relationship between two variables—the covariance and correlation coefficient. The tools of descriptive analytics are helpful in understanding the data, identifying the trend or patterns in the data, and making sense from the data contained in the databases of companies. The understanding of databases, data warehouse, web search and query, and big data concepts are important in extracting and applying descriptive analytics tools. The flow chart in Figure 4.1 outlines the tools and methods used in descriptive analytics.

Figure 4.1 Tools and methods of descriptive analytics

Descriptive Analytics Applications: A Business Analytics Case

A case analysis showing different aspects of descriptive analytics is presented here. The case demonstrates the graphical and numerical analyses performed in an online order database of a retail store and is described below.

Case Study: Buying Pattern of Online Customers in a Large Department Store

The data file “Case-Online Orders.xlsx” contains data on 500 customer orders. The data were collected over a period of several days from the customers placing orders online. As the orders are placed, customer information is recorded in the database. Data on several categorical and numerical values are recorded. The categorical variables shown in the data file are day of the week, time (morning, midday), payment type (credit, debit cards, etc.), region of the country order was placed from, order volume, sale or promotion item, free shipping offer, gender, and customer survey rating. The quantitative variables include order quantity and the dollar value of the order placed or “Total Orders.” Table 4.1 shows the part of the data.

Table 4.1 Partial data online orders

The operations manager of the store wants to understand the buying pattern of the customers by summarizing and displaying the data visually and numerically. He believes that using the descriptive analytics tools including the data visualization tools, numerical methods, graphical displays, dashboards, and tables of collected data can be created to gain more insight into the online order process. They will also provide opportunities for improving the process.

The manager hired an intern and gave her the responsibility to prepare a descriptive analytics summary of the customer data using graphical and numerical tools that can help understand the buying pattern of the customers and help improve the online order process to attract more online customers to the store.

The intern was familiar with one of the tools available in EXCEL—the Pivot Table/Pivot Chart that she thought can be used in extracting information from a large database. In this case, the pivot tables can help break the data down by categories so that useful insight can be obtained. For example, this tool can create a table of orders received by the geographical region or summarize the orders by the day or time of the week. She performed analyses on the data to answer the questions and concerns the manager expressed in the meeting. As part of the analysis, the following graphs, tables, and numerical analyses were performed.

1. 1.A pivot table, a bar chart, and a pie chart of the pivot table providing a summary of number of orders received on each day of the week were created to visually see the orders received by the online department on each day (Figures 4.2 and 4.3). The table and graphs show that the maximum number of orders were received on Saturday and Sunday.

Figure 4.2 Number of orders by day

Figure 4.3 Number and percent of orders

1. 2.Table 4.2 and Figure 4.4 show the count of number of orders by the time of the day (morning, midday, etc.). A bar chart and a pie chart of the pivot table were created to visually see the orders received online by the time of day. The pie chart shows both the numbers and the percent for each category. The table and the pie chart indicate that more orders are placed during night hours.

Table 4.2 Number of orders by time

Figure 4.4 Plot of number of orders by time

1. 3.Orders by the region: The bar chart and the pie chart (Figures 4.5 and 4.6) summarize the number of orders by the region. These plots show that the maximum orders were received from the North and South regions. Marketing efforts are needed to target the other regions.

Figure 4.5 Number of orders by region

Figure 4.6 Percent of orders by region

1. 4.A pivot table (Table 4.3) and a bar graph (Figure 4.7) were created to summarize the customer rating by gender where the row labels show “Gender” and the column labels show the count of “Customer Survey Ratings” (excellent, good, fair, poor). A bar chart of the count of “Customer Survey Ratings” (excellent, good, fair, poor) on the y-axis and gender on the x-axis is shown below the table. This information provided the customer opinion and was important to view and improve the process.

Table 4.3 Customer ratings

Figure 4.7 Customer ratings by gender

1. 5.The descriptive statistics of the “total orders ($)” was calculated and displayed in Table 4.4 and the plot below. The statistics show the measures of central tendency and the measures of variation along with other useful statistics of the total orders.

Table 4.4 Descriptive statistics of total orders

1. 6.From the calculated statistics in part (5), it seems appropriate to conclude that the total orders data are left skewed so that Chebyshev’s rule can be applied. This rule applies to any distribution, symmetrical or skewed, and relates the mean and standard deviation to provide more insight. This rule is too general and does not provide definite conclusions. More definite conclusions can be drawn using the other widely used rule known as the empirical rule that applies to symmetrical or normal distribution. This rule also provides a relationship between the mean and standard deviation of the data and provides a more definite conclusion.
2. 7.If the total orders data can be assumed to be approximately symmetrical, what conclusions can we draw about the “total orders” (Figure 4.8) received? Use the mean and standard deviation calculated in part (5).

Figure 4.8 Graphical summary of the total orders data

A symmetrical or bell-shaped data that is characterized by a normal distribution is often used to draw conclusion by combining the mean and standard deviation. If the data can be approximated by a normal distribution, an empirical rule applies. For our case data, if we can assume that the “total orders” data is approximately symmetrical, we can draw the following conclusions relating the mean and standard deviation of the “total orders” data that were calculated in part (5).

Conclusions using Empirical Rule are shown in Table 4.5.

Table 4.5 Conclusions using the empirical rule

The mean and standard deviation of total orders are:

Variable: Total Orders ($): Mean: = 223.87

Standard Deviation: s = 84.23

1. 8.A dashboard (shown in Figure 4.9) provides several views of the online orders data on one plot. The dashboard below shows several plots, including order map showing the business activities in different regions of the country, sales by months, percent of orders by time, and total orders by region. The plots are self-explanatory and provide useful information that provide opportunities for improvement. The graphical and numerical analyses performed on the online order data provide meaning and insight that is not apparent just by looking into the data. The analyses performed here are some examples of descriptive analytics.

Figure 4.9 A dashboard of online orders data

The type of analytics that goes beyond descriptive analytics is predictive analytics. Applying the tools of descriptive analytics enables one to gain insight and learn from the data. These tools help to understand what has happened in the past and is very helpful in predicting future business outcome. Predictive analytics tools help answer these questions. The rest of the book explores predictive analytics tools and applications.

Summary

In this chapter, we provided a brief description of descriptive analytics and a case to illustrate the tools and applications of visual techniques used in descriptive analytics. The descriptive analytics is critical in studying the current state of the business and to learn what has happened in the past using the company’s data. The knowledge from the descriptive analytics lays a foundation for further analysis and leads to predictive analytics. As mentioned, the knowledge obtained by descriptive analytics helps us to learn what has happened in the past. This information is used to create predictive analytics models.

The subsequent chapters discuss the predictive analytics and background information needed for predictive analytics along with the analytical tools. Specific predictive analytics models and their applications are the topics of chapters that follow The rest of this book covers mostly predictive analytics.