Introduction

Sources such as online behavior tracking software, loyalty programs, scanner systems, online product reviews, and consumer surveys, to name a few, provide businesses with unprecedented access to marketing-relevant data. If utilized effectively, these data can inform and support decision-making across a wide range of marketing functions. Consequently, marketing analytics, or the use of quantitative data to make marketing decisions and evaluate marketing performance, has become critical to the success of many firms. In support of this assertion, results from the recurring CMO survey conducted by Duke University (cmosurvey.org) indicate that US companies’ use of, and spending on, marketing analytics is growing rapidly. The Spring 2023 survey revealed that 77.0% of respondents (314 marketing leaders at for-profit US companies) reported that marketing was responsible for marketing analytics in their companies. Further, when asked “Which activities does your senior marketing leader find challenging to implement on a regular basis,” 41.6% of the 316 respondents in the Fall 2023 survey selected “focusing data and analytics on the most important marketing problems.” This response ranked third most important among the eight response options. The widespread and growing use of marketing analytics is not surprising considering that the use of analytics has been shown to be positively related to both profits and ROI (Ariker et al. 2015). Further, the challenges senior marketing leaders encounter in focusing data and analytics on important problems speaks to employers’ desires to find employees with quantitative job skills (Schlee and Harich 2010; Wilder and Ozgur 2015).

With the (growing) importance of marketing analytics, there is a need to understand how to effectively develop analytics-related skills in students to prepare them for the workforce. The current research addresses this need by (a) using a theory-based approach, scaffolded learning, to develop student comfort with a commonly used analytics tool (Microsoft Excel) and (b) providing educators with resources for doing so (i.e., the exercises used to develop relevant skills and comfort).

Business analytics covers many activities and tasks and requires both technical and business knowledge (e.g., Power et al. 2018; Wilder and Ozgur 2015). The necessary technical knowledge can include front-end functions involving data collection (e.g., installing hardware and/or software needed to build scanner systems or create online surveys). It can also include advanced analysis skills (e.g., building complex predictive models). However, these are highly specialized functions. More commonly, employees need to have basic skills to enable them to access, manipulate, organize, explore, summarize, or visualize data. A useful, and widely used, tool for achieving these goals is Microsoft Excel (hereafter, Excel).

Recent surveys estimate over 80% of jobs require proficiency in Excel or similar software (Adams 2022), 37% of office workers’ time involves using Excel, and 66% of professionals report looking at a spreadsheet at least once an hour (Acuity Training 2022). Despite being on the market for almost 40 years, Excel remains an important decision-making tool across a wide range of organizations and job titles (e.g., Lawson et al. 2009). An analysis of job postings from 2019 to 2021 led Rebman et al. (2023) to conclude, “Results overwhelmingly indicate that Microsoft Excel still is the most required spreadsheet application by employers and faculty should pause before changing MS Excel training or removing certifications.” Consequently, as universities develop analytics-based courses and programs (Pilling et al. 2012; Saber and Foster 2011; Weathers and Aragón 2019; Wilder and Ozgur 2015), fostering Excel skills should be an important component of the curriculum. Not only will doing so equip students with skills employers find attractive, Excel can also effectively support student learning of more general business concepts (e.g., Barreto 2015).

The primary purpose of the current research is to assess the effectiveness of a scaffolded approach for building student comfort with Excel. Scaffolded learning occurs when teachers first develop fundamental concepts in students’ minds. They then contrast these concepts with other ideas, thus building and broadening students’ knowledge structures by allowing students to accommodate the ideas’ unique features in memory (Wood et al. 1976). With data collected across seven sections of an undergraduate Marketing Metrics and Analytics course, we provide evidence that a scaffolded approach does enhance the confidence students have in their Excel skills. This effect is robust across levels of student ability and general performance in the course. To support analytics education, we also provide the Excel exercises used in the course as a resource for instructors.

Literature review

To inform our research, we draw on literature from the areas of scaffolded learning, spreadsheet usage, and student comfort.

Scaffolded learning

Humans learn through a process of scaffolding of ideas (Wood et al. 1976). Scaffolding occurs when more advanced concepts are based on fundamental principles (e.g., Larkin and Beatson 2014). Once fundamental concepts are built in the mind, other ideas can be contrasted and accommodated in memory with their unique features, thereby building and broadening a knowledge structure. While there are various approaches for implementing scaffolded learning, it occurs, in general, through a “process whereby one person in the role of ‘teacher’ mediates the progress of another person, the ‘learner,’ by reducing the scope for failure in the task the learner is attempting” (Maybin et al. 1992, p. 23). The process has been described as “I do. We do. You do.”; that is, the teacher shows how a task should be completed prior to allowing students to work individually to complete the task (Teaching and School Administration 2023).

Scaffolding is not appropriate for all tasks (Maybin et al. 1992; Taber 2018). Simple tasks that can be completed without instructor support can lead to boredom. Extremely difficult tasks that are not possible even with instructor support can lead to frustration. However, moderately difficult tasks, in which success is possible with some instructor support, are amenable to scaffolding. Learning Excel is likely to fall into this category. As noted by Barreto (2015), “We know any student can open a spreadsheet and use rudimentary formulas, but there is enough complexity in achieving advanced proficiency that not everyone can easily master it. On the other hand, advanced proficiency is not so difficult that it becomes a computer science class. Excel is ‘just right,’ hitting the sweet spot between too easy and too hard.” As we subsequently discuss, reported error rates for various features of Excel (and spreadsheets in general) serve to support this assertion.

Spreadsheet usage

Spreadsheets have become engrained as decision-making tools across a range of business applications (e.g., Chan and Storey 1996). As reported by Permberton and Robson (2000), the most used spreadsheet facilities include sorting, database functions, tabulating and organizing, graphical presentation, and summary measures. Advanced statistical analyses, such as correlation and regression, are used less frequently (though their relative use has likely increased in the years since this research was published; see also Chan and Storey 1996). Based on frequency of use, as well as the learning objectives of the Marketing Metrics and Analytics course which served as the basis for this research, Excel training focused on the first five facilities listed but did not include correlation, regression, or more advanced statistical analyses. Table 1 summarizes the specific facilities included in the exercises.

Table 1 Course-relevant excel facilities
Full size table

Despite the importance and widespread use of spreadsheets, it is generally accepted that errors are prevalent. Because errors can lead to suboptimal decisions, multiple taxonomies to classify spreadsheet errors have been developed in hopes of understanding their causes, reducing their prevalence, and mitigating their impact (Powell et al. 2008). For example, Panko and Halverson (1996) distinguished between quantitative and qualitative errors. Quantitative errors lead to wrong numbers in the current spreadsheet, while qualitative errors could lead to errors in subsequent uses, but not the current version, of the spreadsheet. Panko and Halverson (1996) further classified quantitative errors as mechanical (mistakes in typing), logic (choosing the wrong function or creating the wrong formula), or omission (misinterpreting the situation to be modeled). Teo and Tan (1997) built on this classification scheme by adding two categories of qualitative errors: jamming (more than one parameter is placed in a single cell) and duplication (same parameter occurs in two or more cells). Rajalingham et al. (2000) and Purser and Chadwick (2006) developed more elaborate taxonomies in which the highest level distinguishes between application-identified errors (e.g., Excel indicates #DIV/0! when attempting to divide by 0) and user-identified errors (i.e., errors recognized by the developer or user). User-identified errors can be classified as qualitative or quantitative. Quantitative errors include both reasoning errors and accidental errors. Whereas accidental errors arise from an unintended mistake by the user/developer, reasoning errors can arise from a lack of domain knowledge (i.e., lack of real-world knowledge or mathematical knowledge) or from implementation (e.g., problems with syntax or logic).

While our goal is not to classify errors or quantify error rates, for completeness, we note that our efforts most align with reducing quantitative reasoning errors. As discussed, business analytics requires both business and technical knowledge (e.g., Power et al. 2018; Wilder and Ozgur 2015). Because reasoning errors can arise from lack of domain knowledge, the Marketing Metrics and Analytics course is designed to equip students with the real-world and mathematical knowledge to create a spreadsheet to assess a given marketing domain. Further, because implementation errors arise when users/developers lack the skills or logic to correctly build the spreadsheet, the Excel exercises used in the course seek to build implementation skills. For example, the course section on customer metrics informs students about the concept of customer lifetime value (real-world knowledge) and provides a formula for computing customer lifetime value (mathematical knowledge) before students complete an Excel exercise in which they use this formula to determine lifetime value (implementation skills).

The business/marketing impact of any given spreadsheet error can be small or large, depending on the decision at hand. Regardless of the size of the impact, errors compromise decision-making to some degree and, therefore, should be a cause for concern. Powell et al. (2008) provide a summary of research that identified and quantified the types of spreadsheet errors. While identifying and classifying errors and quantifying error rates can be difficult (e.g., Singh et al. 2017), a key takeaway from this summary is that error rates are non-trivial. This supports our assertion that learning Excel is neither extremely difficult nor trivial and, thus, is appropriate for a scaffolded approach.

Student comfort

The goal of developing expertise with Excel lends itself to objective assessment (e.g., error rates). That is, either a user correctly performs a spreadsheet task, such as entering a formula, or the user does not. However, how instructors should go about improving not only students’ learning of such tasks but also their own ability to teach students to do so is less clear. The fundamental thesis of the Scholarship of Teaching and Learning (SoTL) literature is that the best way to do this involves viewing teaching as a topic for scholarly inquiry, being steeped in the literature on teaching, adapting the literature on teaching to discipline-specific content, and taking efforts to improve learning public for the purposes of scrutiny and dissemination (e.g., Crittenden 2023; Trigwell 2021).

At the same time, it is important to remember that learning involves not only skill development and disciplinary knowledge, but also students’ feelings and attitudes about their learning (Felten 2013). Extending SoTL in this direction, we focus on student comfort, as research suggests that a user’s comfort with Excel has important implications. Comfort is a subjective feeling and refers to a lack of tension, apprehension, nervousness, or worry about a task (Horwitz 1986). Comfort has been shown to be positively related to interest in marketing analytics (Weathers and Aragón 2019), learning and student mastery of course content (e.g., Che Ahmad et al. 2017; Dallimore et al. 2010), and student persistence (Gloria et al. 1999). Further, comfort may be enhanced by scaffolded approaches (Hutchison 2016). The converse of comfort, anxiety, is a reliable indicator of students who are at-risk of performing poorly in quantitative courses (Pilling and Nasser 2015). Reducing anxiety can positively impact learning outcomes in areas such as science (e.g., Hong 2010) and languages (e.g., Ganschow and Sparks 1996).

The moderate level of challenge associated with learning Excel (Barreto 2015), coupled with research demonstrating that scaffolded learning approaches can enhance comfort (Hutchison 2016), lead us to predict that a scaffolded approach to teaching Excel skills should be effective for enhancing student comfort with Excel. In the next section, we describe our methodology for testing this prediction.

Methodology

We collected data from seven sections of an undergraduate Marketing Metrics and Analytics course held between Spring 2019 and Summer 2023. Students typically take the course, required for Marketing majors, during their sophomore or junior year. The course description states that the course: “Examines the derivation, meaning, use and communication of marketing metrics used to facilitate decision-making in various areas, including, but not limited to, online and social media strategy, advertising, pricing, branding, and product development. Students are also introduced to database management, including the use of Microsoft Excel.” During the first week of the class, students completed a survey that included a battery of 22 items assessing their comfort with various aspects of the course. All items were measured on seven-point scales anchored by “extremely uncomfortable” and “extremely comfortable” (e.g., Weathers and Aragón 2019). Five of the items were specific to Excel. These items included “Rate your level of comfort as of today with: (1) using Microsoft Excel in general, (2) using Excel to create charts, (3) using PivotTables in Excel, (4) using Excel to manipulate data, and (5) using functions in Excel.” With single items employing seven-point scales, students indicated how confident they were in their ability to do well in the course (not at all confident/very confident) and how important the course was based on their interests and career goals (not at all important/very important). During the final week of the course, after all Excel exercises had been completed, students responded to the same battery of comfort items. In addition, we obtained a measure of general student ability (GPA prior to taking the course), and we recorded each student’s final course average (0–100).

Throughout the semester, students were assigned a series of 14 Excel exercises, summarized in Table 2 and provided in their entirety in the Online Appendix. Each exercise involved students trying to create a chart, table, or other output (e.g., a measure of ROI). The specific output required was provided in the exercise (e.g., a screenshot of the chart), and students were given instructions on how to recreate this output. If students were successful in recreating the specified output, they received credit for the exercise. Otherwise, no credit was awarded. Students were expected to complete the exercises individually, though they could ask the instructor for help. The exercises followed a scaffolding approach. When a new Excel skill was required to successfully complete the exercise, the instructions explained the skill to students in words and images, as illustrated in Fig. 1. Once a skill was explained in detail, students were expected to have learned the skill and were not provided with this level of detail when using the skill in subsequent exercises. The course was structured by different marketing functions, such as pricing, product development, advertising, and social media. The focus of each exercise aligned with one of these functions. As a final project, students used Excel to profile a Netflix user’s viewing behavior (viewing activity by time of day, day of week, month of year, length of program, and device). The data represented either the student’s personal data obtained from Netflix or a hypothetical user’s data provided by the instructor.

Table 2 Purpose of Exercises and Excel Facilities Included
Full size table
Fig. 1
figure 1

Example of instructions provided in exercises

Full size image

Of the 232 students enrolled in the seven sections (average of 33 students per section), 168 completed the comfort measures at period 1 (beginning of semester) and period 2 (end of semester) and are included in the following analyses. An exploratory factor analysis of the 22 comfort items revealed the five measures of comfort with Excel loaded highly on a single factor at both period 1 (α = 0.94) and period 2 (α = 0.91). Thus, we averaged responses to the five items to create an overall measure of comfort with Excel for each period. For completeness, we also present the results for the individual measures.

To assess student ability, we standardized the measure of GPA [(GPA − mean GPA)/standard deviation of GPAs]. We also standardized each student’s final average by section [(student average − class average)/standard deviation of student averages] to account for any differences in grading standards or weights across sections or semesters.

We ran two repeated-measures ANCOVAs, both of which included period as the repeated factor (that is, comfort at period 1 and comfort at period 2). In the first analysis, we included covariates reflecting students’ characteristics as they entered the course, including GPA (ability), confidence in their ability to do well in the course (confidence), and perceived importance of the course based on the interests and career aspirations (importance). In the second analysis, we included as a covariate students’ final average grades, which reflected their performance across the course. Table 3 presents the results for each analysis.

Table 3 ANCOVA results
Full size table

For the first ANCOVA, the main effect of period was significant (F(1, 160) = 300.09, p < 0.01). The confidence × importance (F(1, 160) = 12.54, p < 0.01), period × confidence (F(1, 160) = 3.96, p = 0.05), and ability × confidence × importance (F(1, 160) = 5.05, p = 0.03) interactions also achieved significance. No other main or interactive effects were significant at the 0.05 level (smallest p = 0.09).

For the second ANCOVA, the main effects of period (F(1, 166) = 316.39, p < 0.01) and final average (F(1, 166) = 7.69, p < 0.01) were significant. The period × final average interaction was not significant (F(1, 166) = 0.03, p = 0.86).

The significant (and large) main effect of period in both analyses revealed that students were more comfortable with Excel at the end of the semester (MT2 = 6.27) than at the beginning of the semester (MT1 = 4.30). Paired-samples t-tests revealed this main effect was robust for each measure of comfort with Excel: using Excel in general (MT1 = 4.80, MT2 = 6.38, t(167) = 14.28, p < 0.01); using Excel to create charts (MT1 = 4.77, MT2 = 6.52, t(167) = 14.21, p < 0.01); using PivotTables in Excel (MT1 = 3.77, MT2 = 6.00, t(167) = 15.03, p < 0.01); using Excel to manipulate data (MT1 = 3.84, MT2 = 6.17, t(167) = 18.44, p < 0.01); and using functions in Excel (MT1 = 4.31, MT2 = 6.29, t(167) = 15.96, p < 0.01). Figure 2 illustrates these results. Further, paired-samples t-tests showed that students entered the course with significantly greater comfort in their ability to create charts relative to using PivotTables, manipulating data, and using functions, and they were more comfortable with using functions than with using PivotTables and manipulating data.

Fig. 2
figure 2

Comfort with excel at period 1 and period 2

Full size image

For insight into the significant confidence × importance interaction, we used the method of seemingly unrelated regressions (SUR; Zellner 1962) with Excel comfort at period 1 and Excel comfort at period 2 as the dependent variables and confidence, importance, and their interaction as the independent variables. For period 1 comfort, the main effect of confidence was significant (b = 0.24, std. b = 0.16, p < 0.05), the main effect of importance was not significant (b = 0.12, std. b = 0.08, p = 0.29), and the confidence × importance interaction was significant (b = 0.34, std. b = 0.22, p < 0.01). For period 2 comfort, the main effect of confidence was not significant (b = 0.03, std. b = 0.04, p = 0.59), the main effect of importance was not significant (b = 0.06, std. b = 0.08, p = 0.32), and the confidence × importance interaction was significant (b = 0.17, std. b = 0.20, p < 0.01). The positive confidence × importance interaction on Excel comfort indicates that the positive impact of confidence on Excel comfort was even stronger as students perceived the course as more important.

To understand the significant period × confidence interaction, we computed the difference in Excel comfort between periods 1 and 2 (comfortT2 – comfortT1) and correlated this difference with the measure of confidence. The correlation was positive and significant (r = 0.19, p = 0.01), indicating that students with greater confidence tended to see a larger increase in comfort than students with less confidence. This result is noteworthy since it is contrary to a “ceiling effect” prediction in which students who have greater confidence entering the course would be expected to have high levels of comfort with Excel, leaving little room for an increase in comfort. The result is also contrary to a “boredom” prediction in which students who have greater incoming confidence would feel less challenged by Excel, leaving little motivation to increase comfort. One possible explanation for why students with greater confidence tended to see a larger increase in comfort than students with less confidence is based on research showing that students with more (versus less) prior experience in a domain show greater increases in self-efficacy after additional training in the domain (Lent et al. 2006; Spangler et al. 2014). For the present context, this suggests that students who come into the course with higher (versus lower) levels of generalized confidence may be better positioned to develop specific confidence with Excel through scaffolded learning, resulting in greater comfort with Excel.

For insight into the significant final average main effect in the second ANCOVA, we used the SUR method, this time with Excel comfort at period 1 and Excel comfort at period 2 as the dependent variables and final average as the independent variable. Final average had a marginally significant positive effect on comfort for period 1 (b = 0.26, p = 0.06) and was significantly positively related to Excel comfort at period 2 (r = 0.24, p < 0.01). These results indicate that students who ultimately performed better in the course overall tended to report higher levels of comfort with Excel, both at the beginning and end of the course.

As a final point of inquiry, we examined the relationship between comfort and proficiency with Excel. This analysis informs instructors who may be concerned that some students would report feeling comfortable with Excel despite lower proficiency, whereas other students would report feeling discomfort with Excel despite higher proficiency. To perform the analysis, we measured each student’s proficiency based on their semester grade across all Excel assignments and then correlated it with both measures of comfort with Excel (i.e., comfort at the beginning of the semester and at the end of the semester). Proficiency was not significantly correlated with comfort at the beginning of the semester (r = 0.09, p = 0.27). However, proficiency was significantly positively correlated with comfort at the end of the semester (r = 0.32, p < 0.01). Importantly, we find that the two correlations are significantly different (z = − 2.45, two-tailed p = 0.01; Steiger 1980). We subsequently discuss these findings in more detail.

Conclusions and future research

With this research, we demonstrated the effectiveness of a scaffolded learning approach for enhancing student comfort with Excel. Specifically, students were provided with detailed instructions when they first encountered the need to use an Excel feature in an exercise, while this level of detail was not provided when the feature was needed in subsequent exercises. This guided learning approach led to students feeling significantly greater comfort with Excel at the end of the semester compared to the beginning of the semester. We provided information about the exercises as a resource for instructors teaching Marketing Metrics and Analytics courses (Tables 1 and 2, Online Appendix).

For several reasons, we encourage Marketing faculty and departments to integrate Excel into the curriculum or assess their current delivery of Excel. As surveys suggest, the importance of marketing analytics continues to grow, and Excel continues to be a widely used decision-making tool across a range of companies and business/marketing applications. Consequently, students need to be equipped with skills to operate in data-rich environments. As noted by Permberton and Robson (2000), limited or inappropriate Excel training will result in companies being unable to fully capitalize on the potential of spreadsheets to enhance decision-making, transfer knowledge across the organization, or improve organization performance. Further, while other software may outperform Excel for specific tasks (e.g., Johnson and Berenson 2019), Barreto (2015) argues that Excel has a “just right” quality. It does not require extensive training but provides enough features and functions to support decision-making, and Excel’s gradual learning curve introduces students to quantitative work and opens their eyes to more sophisticated data analysis and software. Finally, Excel can support learning by providing a rich, multisensory experience that allows students to practice and repeat concepts.

Despite noting potential shortcomings, Johnson and Berenson (2019) argued that Excel “should be used for PivotTable demonstration of data exploration through drill down in all introductory business statistics courses to promote conceptual understanding and enhance statistical thinking.” While we agree with this assessment, we also found that students are less comfortable with PivotTables than other commonly used Excel features. Instructors should be mindful of this comfort deficiency when integrating PivotTables in their courses. These findings suggest that students seem to be learning how to create charts, manipulate data and worksheets, and work with functions in other courses or contexts. However, it appears they are less likely to have experience with PivotTables.

While we hope this research supports the efforts of instructors teaching metrics and analytics courses, we also hope it spurs additional research. We focused on student comfort with Excel, as comfort can contribute to learning and student mastery of course content (e.g., Che Ahmad et al. 2017; Dallimore et al. 2010), interest in marketing analytics (Weathers and Aragón 2019), and student persistence (Gloria et al. 1999). However, instructors may wonder whether comfort implies proficiency, especially when some students report comfort with Excel despite low proficiency, while other students report discomfort with Excel despite high proficiency. Since comfort can have different origins, the relevant question for the present investigation is whether enhanced comfort due to a scaffolded learning process is predictive of proficiency with Excel. As presented, results examining the relationship between comfort and proficiency suggest that, while comfort and proficiency are distinct, they are positively related, lending support to the findings of Che Ahmad et al. (2017) as well as Dallimore et al. (2010). We encourage researchers to further examine the relationship between comfort with Excel and other forms of proficiency, perhaps as reflected in time to complete an Excel-based task or optimizing an outcome with respect to time and accuracy.

Instructors may find value in understanding comfort in conjunction with an established spreadsheet error taxonomy (e.g., Panko and Halverson 1996; Purser and Chadwick 2006; Rajalingham et al. 2000; Teo and Tan 1997). For example, it may be that users are likely to recognize the potential for logic errors, thus creating a healthy amount of discomfort that encourages users to take steps to ensure that they are using the correct logic. On the other hand, users may be oblivious to errors due to omission and not feel discomfort. Consequently, they may not pause to take steps to identify and correct such errors.

This research highlighted scaffolded learning as a mechanism leading to comfort in the context of Excel exercises that were typically performed individually by students. However, many instructors may choose to implement such exercises with students working in teams. Prior research finds that team-based learning can encourage greater engagement (Chad 2012), skill acquisition (Laverie 2006), and functional knowledge (Hibbard et al. 2016), but it remains unclear how Excel exercises may be best designed to encourage scaffolded learning in team environments. We encourage future research to examine this possibility.

We used an undergraduate Marketing Metrics and Analytics course as the empirical context for assessing the impact of a scaffolded learning approach on student comfort with Excel. However, MBA and Master of Science in Marketing programs have been equally, or perhaps more, proactive in adapting their curricula to business organizations’ demand for employees who are prepared for working in data-rich environments (Weathers and Swain 2021). Because graduate students are typically older than undergraduate students, and often have work experience (perhaps with Excel), we feel there is value in examining potential differences in scaffolded and other Excel learning approaches across undergraduate and graduate populations.

To these points, a principle of SoTL is the importance of context. As presented by Felten (2013), citing work of Huber and Hutchings (2005), SoTL is “rooted in a particular classroom, disciplinary, institutional and cultural contexts” (p. 122). Felten subsequently states, “any measure of good practice must account for both the scholarly and the local context where the work is being done” (p. 123). Thus, we acknowledge that the results of this research were obtained for a particular class structure, professor, and population of students. However, it is our hope that the exercises and scaffolded approach used provide a starting point for instructors seeking to build student comfort with Excel.

As a final point of future inquiry, while research cited from other areas of business provides some insight into the most used spreadsheet features and facilities, we encourage researchers to examine the rates of usage for various Excel features across marketing functions. Such insight could inform the development of marketing exercises, courses, and curriculum.