1. Introduction
In almost every management area, significant quantities of data are generated and used in digital formats. Intrinsically business analytics can be defined as competencies, tools, methods for ongoing iterative analysis and review into previous company results in order to obtain the insight and improve the strategic planning and helps in deciding which databases are useful and how to manipulate them to solve challenges and improve performance, competitiveness, and sales (Laursen & Thorlund, 2016). Moreover, efficient business analytics depends on data reliability, technical and market understanding by professional analysts, and organizational commitment that use analytics to derive insights to facilitate strategic decisions.
In a statement by Vidgen et al. (2017), when the market objective of the analysis is identified, an analysis methodology is formulated, and information is complied to validate the research. Data management often comprises downloading, processing, and updating to regular service, including a centralized database, and thus computational approaches vary from realistic mathematical table sheets to sophisticated data analysis and predictive simulation applications (Vidgen et al., 2017). Therefore, business analytics is a more prescriptive method in which data can be collected, patterns recognized, and models designed to describe recent events, forecast future events and suggests steps to maximize the outcomes.
The current report will illustrate potential data-driven techniques to the selected data to address problems related to managerial decision making, which helps the company executives or managers grasp their industry environment, predict demand trends, and mitigate risks. The sections below will portray the three main problems in the current datasets, which are sales analysis and forecasting, customer engagement insights, and sales classification of fast-moving consumer goods and the data-driven techniques which can be implemented to solve these three problems.
1.1 Supermarket Dataset
The dataset is one of the supermarket company historical sales and reported data for three months in 3 separate branches. For the study, the supermarket sales data is selected as the growth of supermarkets is growing in most populated cities, and market competitions are also high accordingly. The data consists of 17 attributes out of which branch, city, customer type, gender, product line, quantity, date, time and payment methods are the most crucial which can help the managers to analyze the sales forecasting, customer engagement insights, and rapid-moving consumer products to improve the sales challenges using the business analytics. The following figure shows the screenshot of the selected datasets:
Figure 1 – Supermarket datasets (Curated By the author from Kaggle website)
The current report will also illustrate how the three data-driven techniques can be implemented in business analytics software such as Microsoft Excel and RStudio, which is a spreadsheet and programming language that can assist in statistical computation and graphics used to implicate the result before the final recommendation and conclusion.
2. Three problems related to supermarket sales datasets Business analytics advancement allows the retailers to draw on large volumes of consumer data, which helps them to develop customized products and contribute to maintaining the customer loyalty through consumer data mining (Hofacker et al., 2016). However, this section will portray the three different sales problems in the current selected supermarket sales datasets:
2.1 Sales Analysis and Forecasting
Sales analysis and forecasting is the method of predicting potential demand where reliable revenue predictions enable businesses to make smarter business decisions and forecast performance for both the short and the long term (Fan et al., 2017). It offers insight into how a firm can handle its workforce, cash flow, and finances and allows the retailers to minimize the impact by taking advanced steps such as cutting expenses or reorienting the market activities.
However, the challenging part of the present dataset is to analyze the customer-specific data like based on gender or customer type, branch-specific data, and item-related data as all the information is distributed randomly and locating particulars in excel from data providers takes much time, translating them, and storing all of the data in a spreadsheet. Therefore, with all that buzz providing precise measurements and interpretation without expensive mistakes is complicated, and the methods and processes for sales forecasting are prone to error, and the results are often unsatisfactory. Therefore, the regression analysis technique in excel will be used to solve this issue, which can appropriately align customer-specific data to find the sales forecasting.
When the supermarkets build the ideal customer profile, they need to dig further into the interests, expectations, and possible challenges with a product or service of that ideal customer. This helps them to develop and build sales opportunities that target the customer they want. However, retailers need to understand different consumer bases (Carlberg, 2016).
According to Ostrow (2015), most businesses face difficulties in achieving precise and practical revenue projections as they lack clarity in their predicted revenues, and they use inaccurate institution-based forecasting methods that lead to poor outcomes in achieving the sales quota. As a result, data precision is skewed in the business department, and production forecasts are far away from the target. Therefore, data analytics is required to understand the customer expectations for the precise sales forecasting (Abbas et al., 2017).
2.2 Customer Engagement Insights
Gaining market knowledge is essential to understand the target audiences and engaging efficiently with customers (Kunz et al., 2017). This may be extended to a wide range of situations, from beginning a new enterprise or product field to repositioning an existing business or product, from daily planning of potential developments in the market development. To use the interaction as the critical measure of conduct, the companies need to be able to connect the customer experience points with the changes to the bottom line.
The current datasets do not provide any information related to customer segmentation such as high, medium, and low valued customers based on gender or location, which can be beneficial in customer engagement. According to Sawhney et al. (2017), customer segmentation is the process of segregating a company's customers into categories representing customer consistency within each category. Presently, customer segmentation is not possible as the customer data processing and maintenance is hugely time-consuming and opens the door to human errors. The ability to collaborate in excel is also pretty limited as it does not permit multiple users to function concurrently in the same row and does not explain anything about a specific customer upsell possibilities, or whether to give them the follow-up call. Therefore, clustering data-driven techniques in RStudio can be used to mitigate this problem.
The customer segmentation objective is to determine whether to appeal to consumers in each segment to increase each customer's interest in the company. By using consumer segmentation to assess actionable consumer experiments, the marketer can check different strategies on customer focus segments, which are especially important. The outcomes of these promotions can be calculated over time and compared to identify the most successful reproducible deals for each consumer group (Sawhney et al., 2017). For example, concerning the current supermarket datasets, if a retail consumer group is established as 'health and beauty' fans, the marketer might seek a range of upselling strategies to persuade consumers to purchase a new and more costly health and beauty products.
Even though the goal of the marketer is typically to optimize the benefit (income or profit) of each consumer, it is essential to know in advance how any specific marketing activity will impact the consumer which is not possible with the current datasets as the customer is not segmented correctly. As a result, the managers can make choices based on old knowledge leading to disappointed consumers and bad returns, leading to inefficient and unsuccessful marketing processes.
2.3 Sales Classification – Fast-moving consumer goods (FMCG)
According to Dost et al. (2019), FMCG has a low shelf-life due to high competition from the customers or because they are perishable. Moreover, these items are often imported, consumed easily, priced low, and sold in large quantities and even have a high turnover while on shop shelves. In the current dataset, health and beauty products, electronic accessories, home, and lifestyle, sports, and travel are considered to be consumer goods but complex to know which are rapidmoving customer products.
In the FMCG market, the data is highly relevant because of the weak profit margins in maintaining the consumers and attracting the new business, and in order to do so, these firms need to consider and reach their audience accordingly (Pourhejazy et al., 2019). As such, retail analytics provide comprehensive customer experience as well as visibility into the organization's operations and procedures of reach and potential for change. However, there is no information related to sales classification in the current sales datasets, like which product line is fast-moving as all the data is distributed arbitrarily. Intrinsically, the supermarket will be insufficient to know the fast-moving consumer goods, which can dissatisfy the consumers in case if the demand is not fulfilled. Therefore, visual and time-series analysis can be implemented in excel to solve the current issue.
In a statement by Roy et al. (2018), data analytics can provide much more insight than just inventory rates and the success of various items. Moreover, the analytics will also assess how easily advertised items leave the market and foresee where they will need to be replenished, leading to less empty shelves and unhappy customers. Furthermore, this research can also be performed when a sale is underway, by monitoring the first few hours of transactions, which can get a better understanding of how individual products are going, helping the supermarket to forecast the expected stock rates more precisely (Baardman et al., 2017).
3. Data-Driven Techniques to solve supermarket data three problems
The data-driven techniques are defined as the advancement in an operation which is compelled by data rather than by the personal knowledge or perception (Jack et al., 2018). Furthermore, it encourages the organization to analyze and coordinate their data to represent their customers and consumers best. The following points will describe some of the data-driven techniques, which can be used in the current report to solve the sales datasets issues.
3.1 Regression Analysis to solve sales analysis and forecasting
Regression analysis is an aggregation of mathematical techniques used to forecast the relationships between a control dependency and one or more independent variables (Akbulut, 2019). Additionally, this technique is used to forecast model time series and evaluate the relation between the causal effects of variables. Regression analysis also helps one to analyze the impact of observable factors on different scales, such as the influence of price increases and the number of advertising events. Such advantages assist the market analysis, business analyst in extracting, and determining the right selection of variables to be used to construct statistical models (Klimberg et al., 2017).
As such, the sales analysis and forecasting issue can easily be solved using the regression analysis. Companies are continually seeking to improve their statistical capability to get the jump over their rivals. A reasonable forecasting model, for example, helps suppliers to precisely keep the right inventory to meet their product demand (Klimberg et al., 2017). According to Ma et al. (2016), the regression analysis offers guidance in evaluating the economic measures with the most predictive potential. The regression will approximate the influence that a series of independent variables has on another (dependent) variable, by quantifying a single equation.
Currently, the significance of regression analysis is that it allows everybody to assess which factors are necessarily the most, which variables can be overlooked, and how those factors incorporate. The benefits of regression analysis in sales forecasting are that it will ultimately encourage the marketers to crunch the number and help them to make smarter choices about the company at present and in the future (Sagaert et al., 2018). Moreover, it can be employed to forecast the shortand long-term revenue, comprehend stock levels, identify the supply and demand, and analyze and consider how all of these factors affect various variables.
3.2 Cluster Analysis to solve customer engagement insight issue
Clustering, also called cluster analysis, is a crucial subject in data mining. According to Romesburg (2004), the clustering function is to group objects collection in such a way that the objects in the same category are more identical than objects in other classes. Moreover, cluster analysis is an exploratory study that attempts to identify data structures and often referred to as segmentation or the taxonomy study. More precisely, if the classification is not historically established, it aims to classify homogeneous classes of instances.
Consequently, customer engagement insights problem can be solved by using cluster analysis. Cluster analysis is the use of the statistical model in the sense of consumer segmentation to discover clusters of related customers, focused on discovering the smallest differences within customers within each category (Tsiptsis & Chorianopoulos, 2011). Furthermore, the marketing cluster analysis objective is to classify customers to achieve more correctly to achieve more efficient customer targeting through personalization. Currently, three clusters are selected to get the low, medium, and high-value customers concerning gender, location, or branch.
In a statement by Tsiptsis and Chorianopoulos (2011), a common form of cluster analysis is a mathematical algorithm known as a cluster analysis of K-means, often referred to as accurate segmentation. The 'K' cluster points, which are to be centroids, are positioned among the data points in the space. The centroids for which the distance is least is allocated to each data point. Centroids of the new classes are re-calculated after each data object has been allocated. The above two steps are repeated until the centroid movement ceases. This means that the subjective function of having the least squared error is achieved and cannot be further changed. Hence it results in the K cluster (Tripathi et al., 2018).
However, the K-means algorithm struggle as the geometric shapes of the cluster deviates from the spherical shapes and often does not know the number of clusters from the data and needs predefining (Tripathi et al., 2018).
When the advertisers in the shop have a good understanding of the various consumers, they will respond individually to each customer, with the marketing experiences more critical to the particular needs of each client (Flath et al., 2012).
3.3 Sales Trend Analysis and Visual Analysis
Sales trend analysis is the analysis of the past turnover figures to find the trends. Moreover, it is a valuable form of budgeting and financial forecasting that may show the beginning of improvements in the company's near-term revenue growth levels (Cha et al., 2019). In addition to this, a micro-trend for a single commodity could continue for a week, while a macro trend could last for the quarter across the variety of goods. Sales analysis is a convenient way to track progress towards the current sales target while recognizing sales patterns in particular products, consumers, or geographies. Therefore, sales trend analysis can be conducted to solve the sales classification issue.
Moreover, another approach to solving the sales classification issue is by using visual analysis, which consists of a graph, pie chart, and histograms. Visual analytics is an outgrowth of the fields of knowledge visualization, and scientific visualization focused on critical thought for interactive data interfaces (Chaudhary & Murala, 2017).
4. Justifications and Recommendations
The following points will describe the implementation of the above data-driven techniques to solve the sales challenges using business analytics:
4.1 Implementation of Regression Analysis
The following data shows the outcome of regression analysis when implemented in excel to solve the sales analysis and forecasting problem:
From the above figure, Multiple R describes the correlation coefficient and can range between -1 (intense negative relationship) and 1 (strong positive relationship), the absolute value of which shows the intensity of the relation. As the value lies between these, it clearly shows that the variable moves in perfect tandem and in the same direction, and the data is linear (Berger, 2007). The following table shows the ANOVA part, which is used for simple linear regression analysis:
According to Barreto and Howland (2006), the importance of Significance F indicates how accurate the tests are (statistically significant). Moreover, the model is elegant if the value is less than 0.05 (5%). As per the statement by Kraha et al. (2012), regression analysis is useful for the researchers to confront this multicollinear reality, and regression analysis techniques can identify the link between various variables by uncovering trends that were previously unnoticed.
As reported by Perichiappan Perichappan et al. (2018), companies like KPMG uses regression analysis as it provides the right predictors given the associations between the account like sales and its-predefined predictors like sales cost.
4.2 Implementation of Cluster analysis
The following data reveals the effects of cluster analysis when applied in RStudio to address the customer engagement insight issue:
The above figure represents the outcome of customer segmentation based on high, medium, and low valued customers using K algorithm clustering analysis. According to Öner and Öztayşi (2018), by using cluster analysis, the business analysts can analyze the individual customer trends or goods in more depth so that potential campaign will optimize the outcomes.
The below figure is the representation of the histogram of cluster analysis.
4.3 Implementation of Sales trend analysis and Visual analysis
Power Bi software is used to analyze the current supermarket dataset by using sales trends and visual analysis data-driven techniques to know the fast-moving consumer goods. The figure below depicts the implementation of sales trend analysis and visual analysis:
The above figure shows the sales graph, which is a visual contrast overall, which makes it much more digestible at a glance. This can give managers an immediate visual analysis of their performance. Moreover, it not only has reported whether the results are on track or not, but it has also described a pattern review as to whether the latest output pattern is positive or negative overall.
According to Cheriyan et al. (2018), sales trend analysis is a reliable form of budgeting and financial forecasting that may show the onset of improvements in a company's near-term revenue growth levels.
5. Conclusion
The current report discussed the three main problems with the selected supermarket dataset and the justification and implementation of three data-driven techniques to solve those three problems. Among the most crucial aspects of digital marketing is the ability to develop the data-driven insights which direct the marketing campaign. Intrinsically, data is the overwhelming array of analytics and statistics that allow large and small businesses to detect essential patterns and developments for their markets. Additionally, the data works solely to help marketers engage with their customers and their behaviors that enable them to adapt the buyer's experience to the prospective buyers much more. It can be inferred that data is an integral part of effective marketing strategies.
References:-
Abbas, O., Shahzad, K., Ali, G., & Sarwar, U. (2017). Issues on Sales Forecasting for Apparel
Industry. https://www.semanticscholar.org/paper/Issues-on-Sales-Forecasting-for-Apparel-Industry-Abbas-Shahzad/60b3c22ac3bf2af80880d135a85ed8d91915b6f6
Akbulut, A. (2019). Benchmarking of Regression Algorithms and Time Series Analysis Techniques for
Sales Forecasting. http://acikerisim.iku.edu.tr/handle/11413/4844
Baardman, L., Levin, I., Perakis, G., & Singhvi, D. (2017). Leveraging Comparables for New Product
Sales Forecasting (SSRN Scholarly Paper ID 3086237). Social Science Research Network. https://doi.org/10.2139/ssrn.3086237
Barreto, H., & Howland, F. (2006). Introductory Econometrics: Using Monte Carlo Simulation with Microsoft Excel. Cambridge University Press.