Data-Driven Insights into Coupon Usage at Health Foods Case Study Sample

Introduction To Data Analysis of Implementing Hierarchical Clustering, Followed by k-means Clustering for Health Food

The expansion and prosperity of Health Foods, a small grocery store chitin in the United States, depend heavily on its ability to comprehend the tastes and buying patterns of its customers. Using marketing analytics to mine survey data for insights can yield important information for strategic decision-making in today's aggressive industry. This study attempts to investigate coupon utilisation among Health Foods consumers and pinpoint important variables that influence coupon usage. Coupons have a big impact on customer behavior, including purchasing decisions and brand loyalty. This research looks for patterns and trends in the use of coupons among Health Foods' customer base by examining survey data gathered over a four-week period from 1404 consumers.

The research questions can be “Which customers are using Health foods coupon and why?” This subject looks at the behavioral and demographic behaviors of coupon consumers as well as the underlying reasons for their use of coupons. Knowing which consumer groups are more responsive to discounts and why they behave that way might yield useful information for focused advertising campaigns. Health Foods may use this research to guide their coupon distribution strategy and more efficiently spend resources to increase consumer engagement and coupon redemption rates. Furthermore, Health Foods may customize its promotions and message to appeal to various client demographics by understanding the reasons for coupon utilization, which will eventually increase sales and brand loyalty.

This information is significant since it may help shape focused marketing tactics like coupon distribution and advertising campaigns. Health Foods may enhance its customer engagement and retention through couponing by identifying the consumers who are most likely to utilize them and the underlying reasons for their utilization. Preprocessing the survey data as part of the procedures used in this study will guarantee correctness and dependability. The study will employ several statistical methodologies, including regression analysis and clustering algorithms, to ascertain the factors that predict coupon utilization and to categorize clients according to their couponing habits.

Methodologies

Data used

The raw survey data that was gathered from Health Foods consumers over a four-week period is comprised in the dataset that was used for this research. Every entry relates to a unique customer week and records information on shopping habits, such as the amount spent on groceries during that week (Balassa, 2022). The dataset has 1404 instances in total, with four rows assigned to each consumer.

Data Preprocessing

• Handling missing data: The first is filling in the gaps in the data, which is an important part of maintaining the accuracy of the analyses that follow. Missing value instances will not be included in the dataset. This step is crucial to avoiding any biases in the clustering results that might arise from missing data.
• Removing outliers: Extreme outliers in the dataset, which might skew distance computations during clustering procedures, will be closely examined (Bu et al., 2020). To preserve the accuracy of the clustering findings, outliers will be found and removed from the dataset using statistical techniques like z-scores or boxplots.
• Standardizing Scale: It is necessary for variables to have similar scales in order for clustering to work. All variables will be standardized in order to do this, yielding a mean of 0 and a standard deviation of 1. Because Euclidean distance, a popular metric in clustering algorithms, is sensitive to scale changes, standardizing variables is especially important.
• Feature selection: In order to improve the stability of the clustering results, variables that are superfluous or redundant will be eliminated from the dataset. Noise in the data will be reduced by keeping just relevant characteristics, allowing for more precise grouping outcomes (Chikodili et al., 2020).

Hierarchical clustering

• Linkage method: The dataset will be subjected to hierarchical clustering using a variety of linkage techniques, such as Ward's approach, average, and complete. All linking methods have different benefits and drawbacks, thus it is necessary to compare them in order to choose the best one biased on the features of the dataset and the clustering goals.
• Cohenetic correlation: The cohenetic correlation coefficient will be calculated in order to evaluate the hierarchical clustering dendrogram's reliability with respect to the initial pairwise distance matrix (Karna & Gilbert, 2022). An enhanced cophenetic correlation coefficient signifies a more precise depiction of the fundamental data stricture, thus enabling well-informed determination of the ideal number of clusters.
• Dendogram cutoff: To define the appropriate number of clusters, a cutoff point will be found rather than completely splitting the dendrogram (Karthikeyan et al., 2020). This strategic strategy guarantees alignment with corporate objectives and allows for flexibility in cloister allocation depending on business imperatives.

K-means clustering

• Random restarts: Multiple random restarts will be used while executing k-means clustering to reduce the possibility of convergent to poor solutions (Kastner & Fischer, 2023). Starting the clustering method from different starting points for the clusters increases the probability of finding the global optimum and improves the suitability of the clustering results.
• Local optima: Even with random restarts, local optima may occur in k-means clustering. In order to overcome this difficulty, cluster centroids will be initialized by hierarchical pre-clustering. By giving future k-means iterations a better foundation, this preliminary phase hopes to accelerate convergence toward better clustering solutions.
• Selection of K: In k-means clustering, figuring out the right number of clusters (k) is essential. The elbow plot, coefficient, and information criteria are some of the techniques that will be used to assess the quality of the cloistering and determine the ideal number of clusters (Kaur & Kumar, 2022). Moreover, business judgment will be included to guarantee that the chosen k is in line with company goals and practical factors.

Cluster validation

The outline coefficient, one of the cluster validation metrics, will be used to measure how effective the clustering results are. These metrics provide information about how compact and unique a cluster is, which makes it easier to evaluate the quality of the cloistering process (Kent & Schiavon, 2023).

Get Extra 10% OFF on your WhatsApp order!

use my discount

Following these methodological guidelines, our goal is to preprocess the raw survey data and use k-means and hierarchical clustering approaches to identify relevant consumer categories based on their preferences and shopping patterns at Health Foods. These understandings will serve as the foundation for focused marketing campaigns and analytical decision-making procedures, accelerating the business's trajectory toward long-term expansion and diversification.

Results and Limitations

Several criteria stand out as possible predictors for coupon utilization at Health Foods based on the survey data that was presented. Coupon use may be greatly impacted by factors including shopping habits, coupon value, and vegetarian status (Lakens, 2022). Vegetarian customers may be more likely to use coupons for healthier selections. Furthermore, people who stipend cheaply can be more receptive to coupon offers (Kurpjuweit et al., 2021). By examining these factors in conjunction with demographic information, discount campaigns may be targeted to certain consumer categories, which may boost customer loyalty and engagement. In order to properly customize coupon offerings and maximize their influence on sallies and consumer happiness, additional study may include segmentation approaches.

The distribution of data, such as the amount spent at Health Foods and the use of coupons, may be graphically represented by creating boxplots (Mukhametzyanov, 2023). These boxplots would help in spotting trends and gaining a more thorough knowledge of client behavior for focused marketing tactics by offering insights into the central tendency, spread, and existence of outliers within each variable.

The data supplied shows the findings of a principal component analysis (PCA), which was done in an effort to figure out the fundamental structure of Health Foods' customer-related variables. Six components together account for 60.334% of the total variation explained, according to the total variance explained table (Raj & Mohanasundaram, 2020). The links between the original variables and the found components are showing in the rotated component matrix. One such example of a broad customer profile component is the first component, which appears to have a significant relationship with customer-related parameters like Store ID and Customer ID (Serra, 2022). The factors pertaining to shop features, such as store organization and size, are reflected in the second component. Week and Sequence have storing loadings in Component 3, which may indicate shopping habit.

According to a more thorough reading, components four, five, and six may stand for the amount spent, the use of coupons, and the vegetarian status, respectively. The original variables' contributions to each component are displayed in the component transformation matrix.

To help with focused marketing initiatives, this study offers insights into possible client categories and their attributes. Customers with high loadings on component three may be mire impacted by weekly specials, but those with component one characteristics may be loyal or regular shoppers (Sumbherwal et al., 2023). Health Foods may better target their marketing efforts by providing individualized discounts to regular customers or highlighting vegetarian alternatives to those with high loadings on component two by having a better understanding of these underlying structures. Furthermore, determining the variables that contribute to each component directs future attempts to gather data and helps the creation of consumer behavior prediction models.

The eigenvalues of each main component are depicted in the scree plot, which sheds light on their relative significance in explaining the variation in the data. After the first few components in this analysis, the scree plot shows a rapid fall in eigenvalues, indicating decreasing reward is in terms of explaining extra variation (Tarekegn et al., 2020). This highlights how important it is to concentrate on the initial few major components in order to have a meaningful interpretation.

Figure 1: K-means cluster analysis (Initial cluster center and iteration history

(Source: Self-created in SPSS)

The K-means cluster analysis was designed to separate out different client categories according to how they purchase at Health Foods. Convergence was reached after a number of repetitions, suggesting stabile cluster centers. Two separate groupings may be seen in the final cloister centers. Customers who usually shop at medium-sized establishments with a moodiest degree of organization are represented by Cluster 1. They tend to be non-vegetarian, masculine, and have a constant shopping style. Additionally, they spend an average of $128.02 and have a tendency to utilize coupons frequently, favoring higher-value coupons. Customizers that shop at medium-sized businesses, but with better store organization, make up Cluster 2.

They have a varied buying style, are primarily female, and are not vegetarians. Although they also utilize coupons, they often use lowlier-value coupons and spend $60.84 less overall. By using these data to develop marketing tactics that are specifically suited to the interests and behaviors of each cluster, consumer engagement and happiness may be increased.

Figure 2: K-means cluster analysis (Final cluster centers and Annona)

(Source: Self-created in SPSS)

Based on their Health Foods purchase habits, two unique client categories were identified by the K-means cluster analysis. Customizers in Cluster 1 are represented by medium-sized retailers with a modest level of shop organization. They have a dependable buying habit and are more likely to use coupons preferring ones with greater value. Additionally, this group spends an average of $128.02, which is much greater. Customizers in Cluster 2, on the other hand, still shop at medium-sized establishments, but they are more organized. Their purchasing habits are more varied, they utilize coupons less frequently usually choosing lower-value coupons and they spend less overall ($60.84).
Significant variations between the clusters are shown by the ANOVA findings for all variables, indicating that the segmentation effectively captures important changes in consumer behavior. Some factors show significant variance between the clusters, such as gender, amount spent, value of coupons used, and utilization of coupons (Visalia & Gomes, 2020). These variables all have high F-values.

It's crucial to remember that although if the F-tests reveal information about the variations between clusters, their intended application should be descriptive rather than inferential. This implies that the observed significance levels are not adjusted for this and cannot be considered as tests of the hypothesis that the cluster means are equal. The clusters were selected in order to maximize differences across instances. Nevertheless, these results provide insightful information for customized marketing plans that cater to the tastes and habits of every consumer group, thereby raising customer happiness and engagement levels at Health Foods.

Conclusion

This study examined the unique characteristics of coupon use aiming customers of Health Foods with the goal of identifying the underlying variables affecting their usage habits. Through a four-week period of survey data analysis from 1404 consumers, the study aimed to address the following research question: "Which customers are using Health Foods coupons and why?" The results clarified important variables that predicted the use of coupons, such as purchasing patterns, the value of the coupons, and dietary preferences like being a vegetarian. The study used techniques such as hierarchical and K-means clustering to identify different customer categories according to their purchase habits and coupon usage behaviors. These data are crucial for creating marketing efforts that are specifically targeted to the tastiest and habits of each customizer segment.

Significant disparities across consumer groups were also found in the analysis across a number of factors, demonstrating how well the segmentation captured important variances in consumer behavior. Given that the cloisters were specifically chosen to maximize differences, it is crucial to interpret the ANOVA finding is descriptively rather than inferentially, even if they offer insightful information about the differences between the clusters. Still, these results provide Health Foods with useful information to improve customer involvement, optimize coupon distribution, and boost coupon redemption rates. All things considered, this study emphasizes how crucial it is to use marketing analytics to comprehend customer behavior in the cutthroat market of today. Businesses like Health Foods may obtain deep insights into customer preferences and habits by using sophisticated statistical methodology and preprocessing techniques on survey data. This allows them to make well-informed strategic decisions that will fuel long-term development and success.

References

Journals

• Balassa, C. (2022). Recombinator-k-means: an evolutionary algorithm that exploits k-means++ for recombination. IEEE Transactions on Evolutionary Computation, 26(5), 991-1003. Retrieved Form: https://arxiv.org/pdf/1905.00531 Retrieved On [06/02/2024]
• Bu, J., Liu, W., Pan, Z., & Ling, K. (2020). Comparative study of hydrochemical classification based on different hierarchical cluster analysis methods. International journal of environmental research and public health, 17(24), 9515. Retrieved Form: https://www.mdpi.com/1660-4601/17/24/9515/pdf Retrieved On [06/02/2024]
• Chikodili, N. B., Abdulmalik, M. D., Abisoye, O. A., & Bashir, S. A. (2020, November). Outlier detection in multivariate time series data using a fusion of K-medoid, standardized euclidean distance and Z-score. In International Conference on Information and Communication Technology and Applications (pp. 259-271). Cham: Springer International Publishing. Retrieved Form: http://repository.futminna.edu.ng:8080/jspui/bitstream/123456789/5111/1/12_Outlier%20Detection%20in%20Multivariate%20Time%20Series%20Data%20Using%20a%20Fusion%20of%20K-Medoid%2C%20Standardized%20Euclidean%20Distance%20and%20Z-Score.pdf Retrieved On [06/02/2024]
• Karna, A., & Gilbert, K. (2022). Automatic identification of the number of clusters in hierarchical clustering. Neural Computing and Applications, 34(1), 119-134. Retrieved Form: https://upcommons.upc.edu/bitstream/handle/2117/385820/1Automatic%2Bidentification%2Bof%2Bthe%2Bnumber%2Bof%2Bclusters%2Bin%2Bhierarchical.pdf?sequence=3 Retrieved On [06/02/2024]
• Karthikeyan, B., George, D. J., Manikandan, G., & Thomas, T. (2020). A comparative study on k-means clustering and agglomerative hierarchical clustering. International Journal of Emerging Trends in Engineering Research, 8(5). Retrieved Form: https://www.academia.edu/download/63554296/ijeter2085202020200607-61828-1iexml1.pdf Retrieved On [06/02/2024]
• Kastner, J., & Fischer, P. M. (2023). Detecting and analyzing fine-grained user roles in social media?. Computer Science and Information Systems, (00), 6-6. Retrieved Form: https://doiserbia.nb.rs/ft.aspx?id=1820-02142300006K Retrieved On [06/02/2024]
• Kaur, A., & Kumar, Y. (2022). Design of Single and Multiobjective Metaheuristic Algorithms for Effective Data Clustering (Doctoral dissertation, Jaypee University of Information Technology, Solan, HP). Retrieved Form: http://www.ir.juit.ac.in:8080/jspui/bitstream/123456789/8708/1/PHD0258_ARVINDER%20KAUR_196202_CSE_2022.pdf Retrieved On [06/02/2024]
• Kent, M. G., & Schiavon, S. (2023). Predicting window view preferences using the environmental information criteria. LEUKOS, 19(2), 190-209. Retrieved Form: https://escholarship.org/content/qt7rv6936v/qt7rv6936v.pdf Retrieved On [06/02/2024]
• Kurpjuweit, S., Schmidt, C. G., Klöckner, M., & Wagner, S. M. (2021). Blockchain in additive manufacturing and its impact on supply chains. Journal of Business Logistics, 42(1), 46-70. Retrieved Form: https://www.researchgate.net/profile/Stephan-Wagner-14/publication/336850067_Blockchain_in_Additive_Manufacturing_and_its_Impact_on_Supply_Chains/links/6253da15ef01342066693a4c/Blockchain-in-Additive-Manufacturing-and-its-Impact-on-Supply-Chains.pdf Retrieved On [06/02/2024]
• Lakens, D. (2022). Sample size justification. Collabra: Psychology, 8(1), 33267. Retrieved Form: https://psyarxiv.com/9d3yf/download?format=pdf Retrieved On [06/02/2024]
• Mukhametzyanov, I. (2023). On the conformity of scales of multidimensional normalization: An application for the problems of decision making. Decision Making: Applications in Management and Engineering, 6(1), 399-341. Retrieved Form: https://dmame-journal.org/index.php/dmame/article/download/493/138 Retrieved On [06/02/2024]
• Raj, D. D., & Mohanasundaram, R. (2020). An efficient filter-based feature selection model to identify significant features from high-dimensional microarray data. Arabian Journal for Science and Engineering, 45, 2619-2630. Retrieved Form: https://www.academia.edu/download/91188518/s13369-020-04380-220220918-1-tpkygq.pdf Retrieved On [06/02/2024]
• Serra, M. M. (2022). Analysis and Application of clustering and visualization methods of computed tomography radiomic features to contribute to the characterization of patients with non-metastatic Non-small-cell lung cancer. Retrieved Form: https://openaccess.uoc.edu/bitstream/10609/146167/7/mserra4TFM0622report.pdf Retrieved On [06/02/2024]
• Sumbherwal, N., Hooda, B. K., & Vinit, P. K. (2023). Performance Analysis of Distance Measures for Mixed-Variables Data. Retrieved Form: https://www.researchsquare.com/article/rs-3749138/latest.pdf Retrieved On [06/02/2024]
• Tarekegn, A. N., Michalak, K., & Giacobini, M. (2020). Cross-validation approach to evaluate clustering algorithms: An experimental study using multi-label datasets. SN Computer Science, 1, 1-9. Retrieved Form: https://www.wir.ue.wroc.pl/docstore/download/UEWR4e39cdf28b7b48f4a736763d468441dd/Tarekegn_Michalak_Giacobini_Cross_validation_approach_to_evaluate.pdf Retrieved On [06/02/2024]
• Vysala, A., & Gomes, D. J. (2020). Evaluating and validating cluster results. arXiv preprint arXiv:2007.08034. Retrieved Form: https://arxiv.org/pdf/2007.08034 Retrieved On [06/02/2024]