Case Article—Creating a Brick Empire Through Data Visualization and Analytics

1. Introduction

Over the last half century, the proliferation of information systems and technology applications have fundamentally changed the way that organizations operate and make decisions. No longer do managers and workers have to rely solely on their experience and intuition to make important decisions because they can consult timely and relevant data to guide their evaluation of alternatives. The ability to leverage this data to improve organizational decision making has become an important competency that can provide a competitive advantage in the market, a notion that was first popularized by Davenport (2006).

Many academic programs were quick to capitalize on this trend and incorporate “analytics” into their curricula and nomenclature. The term “business analytics” encompasses a wide range of tools and methods that can be classified into three types—descriptive analytics, predictive analytics, and prescriptive analytics. These types are distinct, yet complementary, and students must have a detailed understanding of the distinction between them as well as the insight to know when to apply each tool in practice. Because the field of analytics incorporates methods and techniques from traditional academic disciplines such as statistics, operations research, and management information systems, some academic programs simply rebranded their existing programs without changing their curricula significantly (Klimberg and McCullough 2013). Other programs took a more proactive approach to purposefully redesign their curricula to incorporate the analytical tools, approaches, and methods required by industry. (See Wilder and Ozgur (2015), Liu and Levin (2018), and Zhang et al. (2020) for examples of innovative analytics courses and curricula.)

Professional organizations such as INFORMS also extended their traditional scope to include analytics, going as far as establishing a definition of analytics as “the application of scientific and mathematical methods to the study and analysis of problems involving complex systems” along with a taxonomy of the three types of analytics: descriptive, predictive, and prescriptive (INFORMS 2022). The organization also established a professional certification known as the Certified Analytics Professional (CAP) in 2013 (Nestler et al. 2012).

This case study is designed to address all three types of analytics within the same underlying environment to help students understand the appropriate type of analysis to apply to a given decision scenario. The remainder of this case article is organized as follows. The next section provides a summary of the case scenario, and the following section discusses the learning objectives. The subsequent two sections offer a review of the relevant literature and suggestions for using the case in the classroom effectively. The article concludes with a discussion of the implementation experiences and a conclusion.

2. Summary Discussion

The case is written as a two-part teaching plan consisting of the main case (A) and an optional extension case (B) that instructors can use in courses that cover all the requisite analytical methods. The case tells the story of Greg McCullough, a father of two young girls, who recently read an online article about LEGO investing. Greg thought this might be a good way to earn some extra money to support his daughters’ LEGO-building hobby, and he wants to learn more about the market and the LEGO sets that are currently available for sale at retail stores.

The main case (A) addresses both descriptive and predictive analytics. First, Greg is interested in understanding the characteristics of LEGO sets that are available in each theme (such as City, Friends, Harry Potter, and Star Wars) and two subthemes (Brickheadz and Juniors). He has collected information about all 614 LEGO sets that were released in 2018 and 2019, which comprises the data provided to students. Specific data fields include number of pieces, box dimensions, and weight. Students must conduct a descriptive analysis of the data by theme and subtheme to tease out interesting characteristics of which Greg should be aware. The instructions given to students are purposefully open-ended to require them to determine the information that would be relevant for the decision maker to know in this scenario.

The second part of the main case (A) focuses on predictive analytics, requiring students to develop a model to estimate the intrinsic value of each LEGO set based on its characteristics relative to the other sets that have been released. They can accomplish this via multiple regression analysis to determine the price LEGO “should” be charging based on this intrinsic value. Once the students determine the predicted price for each set, they can compute the value potential, which represents the difference between the predicted price and the actual retail price. Sets with high value potential are strong targets for investing because the intrinsic value of the set is much larger than the retail price LEGO has established. This is similar to the strategy of value investing whereby investors look for securities that are currently undervalued by the market (Battisti et al. 2019). The case asks students to identify the top two and bottom two sets released in 2019 in each theme, subtheme, and various price ranges as a recommendation for sets to concentrate on or stay away from.

The extension case (B) takes the analysis one step further into the realm of prescriptive analytics. Given the value potentials for each set estimated in the main case (A), students must develop an integer program to provide Greg with a recommendation of which sets to purchase and in what quantities to maximize his total value potential subject to constraints such as budget, shelf space, and mixture limitations to ensure proper diversification. The case also asks students to determine the maximum total value potential for a range of budget amounts and recommend a budget level for Greg to use (if he can afford it, of course!).

3. Learning Objectives

The case has general learning objectives with respect to data management and analysis. The analysis required by different sections of the case (i.e., descriptive analysis, predictive analysis, and prescriptive analysis) is also designed to achieve several specific learning objectives.

3.1. General Learning Objectives

A main learning objective is to give students the opportunity to analyze a “bigger” data set than they usually receive in a homework problem or in other teaching cases. Yap and Drye (2018) lament the challenge of obtaining real-world “big data” data sets for use in the classroom. Although this data set does not meet the traditional standards of “big data” because it only contains 614 records, it requires students to apply visualization and predictive tools to summarize the characteristics of specific LEGO sets and themes on a data set that contains more observations than traditional textbook problems or classroom examples.

Another general learning objective of the case is to provide students with experience in dealing with missing data when conducting analysis. There are approximately 10–20 records in the data set that are missing one or more characteristics such as weight or box dimension. These records are related primarily to sets that were recently released at the time the data set was created from information available on Brickset (www.brickset.com) and Amazon’s website. Students must recognize this missing data especially when they fit predictive models; if they do not either omit these records or interpolate the missing values, software packages could give them errors when they try to fit regression models.

One final general learning objective of the main and extension cases together is to provide students with a comprehensive experience of performing descriptive, predictive, and prescriptive analysis to address specific decision objectives. Completing the two cases takes students step by step through the process of using descriptive analysis to summarize a given scenario, using predictive analysis to estimate relationships between variables, and using the output of a predictive model to prescribe an optimal decision alternative to meet various objectives. Dybvig and Cokins (2017) call for a stronger emphasis in practice on the connection between predictive and predictive analysis. These cases reinforce the appropriate application of each type of analysis in practice and the connection between the different types of analytics.

3.2. Descriptive Analysis Learning Objectives

One major learning objective related to descriptive analysis is for students to choose the appropriate analytical tools and approaches to summarize the data in a meaningful way. The first part of the main case (A) suggests that students evaluate the characteristics of LEGO sets across the various themes and subthemes that are currently offered. The instructions are very open-ended, which requires students to make a decision about which types of descriptive analysis and which data filters would be most appropriate to tell a relevant story about the different themes and subthemes of sets.

Similarly, another descriptive analysis learning objective requires students to summarize and present their analysis to the decision maker in a meaningful way. Instructors are free to request any format they want (e.g., oral/video presentations to the class, written reports, or a combination of the two), but it is important that students are asked to present their findings in a cohesive and convincing way. This is an important skill for all analytics professionals, and it is one that is not often practiced or demonstrated within courses via traditional homework problems or exams.

3.3. Predictive Analysis Learning Objectives

The main predictive analysis learning objective requires students to build regression models and find the best subset of independent variables that predict the retail price LEGO should be charging for the set compared with other sets in their product portfolio. The data set provides characteristics about each LEGO set such as number of pieces, number of minifigures, weight, and box dimensions. Students are required to determine the best combination of these independent variables to predict the retail price that should be charged.

Another learning objective is for students to use their predictive model to predict retail prices for each LEGO set in the data file. From that predicted price, students can then compute the value potential for each set, a measure that is described in the main case (A). In some traditional analytics courses, students spend the bulk of their time fitting regression models and determining whether the model assumptions have been violated; but they do not always understand how to use their recommended model to make predictions that can then be used to inform decisions. This case requires them to take that next step in predictive analysis.

3.4. Prescriptive Analysis Learning Objectives

The main case (A) does not include much prescriptive analysis besides asking students to identify the top two and bottom two sets in terms of value potential with respect to various classifications such as theme, subtheme, or price level. Instructors seeking to cover prescriptive analysis learning objectives should consider using the extension case (B) as well. As discussed in the teaching note, it is possible to use the extension case (B) on its own, but it was written to be used as a companion to the main case (A).

The extension case (B) asks students to formulate and solve an optimization model (specifically, an integer program) that determines the best sets to select to maximize value potential while meeting certain diversification goals. This modeling application gives students additional practice with a model of portfolio selection outside of the traditional publicly traded investment context. It provides students with another example of the flexibility of the application of many classic optimization models outside of their initial canonical scenarios.

Completing the full sequence of cases (A) and (B) allows students to experience many of the requisite steps of a practical business analytics project like they could face as an analyst. The only step that is missing is the data collection stage because the case provides the data to the students. Every other step is there in some form, however, from data cleansing (i.e., handling the missing data values) and descriptive analysis to developing a predictive model and using the results to make an actionable recommendation.

One final possible learning objective related to prescriptive analysis could be to ask students to consider soft goal constraints or multiple objectives when developing their optimization model. The main case libraries do not contain many teaching cases that cover goal programing or multiobjective linear programming, so instructors teaching courses involving those methods may want to use the extension case (B) as an opportunity to have students apply those tools. Detailed suggestions are provided in the teaching note.

4. Literature Review

Business analytics is a multidisciplinary field that is applicable to practically every business discipline and functional area. Researchers have developed several frameworks for organizing and representing the vast methods, processes, and techniques that comprise business analytics. Klimberg and Miori (2010) depict business analytics as Venn diagram combining statistics, quantitative methods, and business intelligence. The intersection of these three disciplines is modeling. Jeyaraj (2019) models business analytics skills and competencies as a process flow diagram characterizing a common sequence of analytical steps in a project—acquisition, preparation, analysis, visualization, and interpretation. Gorman and Klimberg (2014) similarly note that analytics programs typically address all the steps of the problem-solving process. This case study addresses many of the tools and processes specified in these frameworks, which demonstrates its relevance to a comprehensive business analytics curriculum.

The increasing emphasis on analytics in practice has led pedagogical researchers to examine the skills and competencies that are required for students to successfully start their careers in business analytics or data analytics positions and to consider how these skills are best developed through analytics curricula. Business analytics practitioners are required to apply common analytical tools such as statistics, predictive modeling, data visualization, and optimization (Johnson et al. 2020). This case requires students to apply each of these tools appropriately to address different decision-making objectives. Successful analytics practitioners also require various soft skills such as problem solving and written communication (Paul and MacDonald 2020, Stanton and Stanton 2020). LeClair (2018) suggests that analytics programs balance the coverage of technical skills with the development of soft skills to ensure that students are well rounded and able to handle the demanding requirements of the marketplace. This case provides students with an opportunity to hone their problem-solving and communication skills by requiring them to determine the appropriate analytical tool to use to address the scenario in the case and summarizing and explaining the implications of their analysis in the form of a written and/or oral report addressed to the decision maker.

The development of new academic programs focused on business analytics has also drawn the attention of researchers who have examined effective methods for teaching analytics at both the undergraduate and graduate level. Drake (2019) advocates the use of case studies to teach operations research and management science methods (which fall under the scope of business analytics) because they can simulate complex decision scenarios that students will face in practice and they require students to solve complex problems requiring higher-level thinking. Vaziri et al. (2022) suggest the use of case studies to increase student motivation in business analytics courses. Wilder and Ozgur (2015) also recommend that analytics programs use specific scenario examples rather than abstract generalizations, and case studies are one form of these kinds of examples. Case studies also represent one form of experiential learning, which Wilson et al. (2018) include as a pillar in an updated analytics curriculum. Gorman and Klimberg (2014) note that students graduating from analytics programs have a wider scope of analytical tools and perspectives to apply to business scenarios than students coming from discipline-specific programs. This case requires students to use a variety of tools and methods, which increases their ability to apply the appropriate analytical tool to business scenarios in practice.

Although many published teaching cases require students to use one or two specific analytical tools to provide a recommendation for a given scenario (e.g., linear programming-based cases such as Köksalan and Batun (2009), Winch and Yurkiewicz (2014), Rao and Belien (2014), and Shechter (2022)), relatively few teaching cases address all three types of analytics—descriptive, predictive, and prescriptive—as this case does. Gorman (2023) describes one such case that focuses on the analysis of performance issues for a small regional railroad. Gorman’s case provides more instructions to the students about the type of analysis to conduct in each part of the case, and it includes less data for students to analyze than the case discussed in this article does. It also does not require students to handle missing data in various fields within the data set. Barnes and Bjarnadottir (2019) discuss a four-part case that includes all three types of analytics within the context of evaluating baseball players and constructing an effective roster. Huchzermeier et al. (2022) illustrate a comprehensive case involving the design and analysis of a customer loyalty program. The cases by Pachamanova (2015) and Kopcso and Pachamanova (2018) explicitly show the connection between predictive and prescriptive analysis, but there is not much focus on descriptive analysis and data visualization. The prescriptive analysis here follows a newsvendor and revenue management framework instead of integer programming. Shumsky (2009) presents six brief interconnected cases that use predictive and prescriptive analytics in the field of airline revenue management, though also without much discussion of visualization. Several other case studies (Ovchinnikov and Pyshkov 2016, Ovchinnikov 2018) also address all three types of analytics, but these cases are more advanced than the kinds of cases that would be used in typical undergraduate courses. This case is flexible enough to be used in undergraduate or graduate environments, a feature that is further discussed in the next section.

5. Suggestions for Classroom Use

This case was written to provide maximum flexibility to instructors regardless of the level of their courses. The case can effectively be used in an introductory undergraduate setting as well as in an advanced graduate environment. This flexibility is mainly derived from the open-ended description of the required analysis in each part of the case. The case does not denote any specific analytical tool that students should use to analyze the scenario, instead leaving the tool/model selection decision up to the student. Instructors in more introductory settings are encouraged to provide students with supplemental guidance about the specific type of analysis they should perform if the students do not have much analytical experience.

The incorporation of all three types of analytics also enables the case to be used in a variety of different courses and programs. The descriptive analysis portion of the main case (A) could be used in a sophomore-level statistics course. It could also be used in an Excel-focused information systems course to illustrate the application of PivotTables and chart tools. The set of both (A) and (B) cases could be used in an analytics or management science/operations research course. The extension case (B) could be used by itself (perhaps after having students read the main case (A) to acquaint themselves with the scenario) in an optimization course. The data set provided can be analyzed with basic Excel tools and with more sophisticated applications such as Tableau, R, and Python.

If instructors want students to complete all the analysis required in the case, it should be completed outside of class as a homework assignment because it would likely take too much class time to perform in class. The homework assignment structure also allows students time to consider which tools or models to use and to reflect on the meaning of their results and recommendations. Students should be required to summarize their results in a written and/or oral report to build their communication skills.

6. Classroom Experiences and Student Feedback

The case has been used in two summer sections in 2020 and 2021 of an online Prescriptive Analytics and Decision Making course within a MS in Analytics and Information Management program. Both sections were taught by the author of the case. Before taking this course, the students have typically taken at least two other courses in the program, whereas some students have taken as many as eight courses.

The students were able to perform all the analysis in each part of the case, although they completed it with varying degrees of detail. Many students produced visualizations of the data set for the given data fields such as average number of sets per theme, average number of pieces per set in each theme, and so on. Some students, however, only produced one or two visualizations rather than considering all the fields. This is to be expected somewhat due to the open-ended nature of the assignment description in the case itself. If many students performed a limited amount of analysis in their submissions, instructors can use this as an opportunity to emphasize the importance of business analysts putting themselves in the position of the client and thinking about the characteristics of the data they would like to learn about if they were making the decision in the scenario. Some students thought deeper about the situation and produced visualizations based on functions of the given data such as average price per piece as well as a proxy for average density (weight divided by volume). Instructors should highlight this type of analysis in their debrief to the class as examples of this kind of deeper thinking about the given scenario.

Some students also produced visualizations that aggregated the data inappropriately. The most common example of this was creating a graph that showed the sum of the number of pieces or the number of minifigures in sets within a given theme. These measures are skewed by the number of sets released in the theme itself. A better measure to present is the average number of pieces or minifigures in sets within each theme. These issues offer instructors opportunities to emphasize the importance of thinking about the data depicted in a visualization and whether that data makes sense or is relevant to the decision scenario.

The students created many different regression models for the predictive analysis part of the case. Many of these included more independent variables than was necessary because a parsimonious model consisting of only two independent variables (see the teaching note for details) has an adjusted R² of 0.9338, which provides enough precision for a scenario like this. Some students created indicator variables for each month that the sets were released and for each of the themes and subthemes, but these variables do not have much predictive power and should be removed from the model through traditional model selection procedures.

After covering integer programming in the course, students were able to complete the prescriptive analysis for the extension case (B) without much difficulty. The most challenging part of the integer program is modeling the mixture constraints to ensure each theme represents at most a certain percentage of the total sets purchased. This is more naturally modeled as a nonlinear constraint, but it can be made linear by multiplying both sides of the inequality by the denominator. An example from class completed before this case assignment had a similar structure, and most students were able to draw the connection between the constraints and model the mixture constraints for the case correctly as linear inequalities. Some, however, did require a reminder to reference that example from class when building the model for the case assignment.

One of the main learning objectives of the case is to take students through a comprehensive decision-making process using all three types of analytics and to reinforce the appropriate application of each type of analytical tool in practice. To assess the outcome of this learning objective, students were asked a question on the final exam in the course that required them to identify the type of analytics (descriptive, predictive, or prescriptive) each of the following tools represents: linear regression, PivotTables, column charts, and integer programming. They used all these tools to complete this case, and the distinction between the types of analytics was not discussed in class outside of the use of this case. Consequently, students’ abilities to correctly classify each tool with the type of analysis it represents can be attributed mostly, if not entirely, to the completion of this case study. In the Summer 2020 course, 17 of 20 students correctly classified all four tools, and the other three students misclassified only one tool. In the Summer 2021 course, all 21 students correctly classified all four tools. This suggests that the case effectively met the objective of reinforcing the appropriate application of each type of tool in practice.

After completing both the main and extension cases, students completed a brief survey to share their feedback about their experiences with the case. Responses to each question ranged from 5 (strongly agree) to 1 (strongly disagree). The results of this survey are presented in Table 1. Student feedback was very positive overall. The students thought the exercise was valuable (Question 1) and helped them to distinguish between the different types of analytics (Question 2) and to know when to apply each tool in practice (Question 3). They thought the case made them think more than other assignments in the program (Question 4) and thought the assignment was enjoyable (Question 6) and should be continued to be used in the course (Question 5).

Table 1. Mean Results from Student Surveys Conducted After Completing the Case Study

Open-ended responses echoed the quantitative survey results.

• I thought the LEGO case study was one of the best parts of this class.

• I found the assignments to be engaging. Looking back, I did learn a lot going through it.

• Having real-world problems is a great learning experience for students. Having to mimic real world constraints [makes it] much easier to understand their importance instead of just reading them in a typical question. I have always valued teachers that don’t just use multiple choice questions or generic questions for assignments, so this was a very fun project to complete.

• I thought the best use of this case was as mentioned as Question 2. Breaking out the project into different pieces looking at each one separately was a really good way to go through this.

• I thought the LEGO case was a great asset to this course!

7. Concluding Remarks

Organizations will continue to use the massive amount of data they collect to make better decisions. As a result, they value employees with advanced analytical skills and capabilities to use these data more effectively. Academic programs have changed significantly over the past two decades to include more analytics content, and many new programs focusing entirely on analytics have been created. This teaching case is a resource for instructors in a wide variety of academic settings to provide their students with a comprehensive decision-making scenario that encompasses all three types of analytics. Students can hone their data visualization, predictive modeling, and optimization skills while at the same time developing coveted soft skills such as problem solving and communication of their analysis and recommendations.