Customer Data Mining for Lifestyle Segmentation

Introduction and the Research Problem

V.L. Migueis, A.S. Camanho, and Joao Falcao e Cunha proposed a study titled “Customer Data Mining for Lifestyle Segmentation” published in the Journal of Expert Systems with Applications, published by Elsevier in 2012. The study discusses that maintaining a good relationship with the customer base ensures companies gain a competitive advantage. Therefore, market segmentation is essential for companies to develop and maintain loyal customers while promoting increased profits.

Furthermore, the authors suggest that the economic and social changes in Europe led to the transformation of the retail sector. This change has caused the relationship between companies and their customers to have evolved significantly. In the past, the companies maintained a different approach towards increasing sales, emphasizing selling products and services without detailed analysis of the types of customers purchasing their products or services. However, globalization has introduced a wide range of competitors; therefore, attracting new customers has required intense efforts to keep the current customer loyal to the brand. In addition, the continuously evolving social and economic conditions have transformed lifestyles. As a result, customers are likely to accept all the information of the companies and opt for better services of competitors when offered more benefits. Therefore, this scenario has led to companies shifting from product and service-related strategies to customer-centric strategies in recent years. However, maintaining customer loyalty has become the primary strategic goal to establish successful customer-centric strategies.

On the other hand, companies aiming to achieve an edge over competitors continually focus on improving the service level to maintain a good business relationship with customers. Indeed, some companies have invested heavily in maintaining their databases and have collected many customer-related data with several data objects to achieve a complete analysis of the purchasing history and pattern. In contrast, this information obtained from the collected data is rarely integrated into the design aspects of the business functions like marketing campaigns. Besides, most companies do not utilize the available information to aid decision-making. Similarly, such an overwhelming amount of data obtained from each customer has caused information overload, but the lack of utilizing this information has resulted in knowledge starvation and lack of potential growth that the companies can achieve. Although analysts are working on various analyses, they cannot keep pace with the voluminous data and turn them into insights for application purposes at a shorter duration. Therefore, the study discusses the utilization of data mining techniques as a tool for data analysis and the implementation of a clustering technique that enables customer segmentation based on the lifestyle and the type of products purchased by the customers for a thorough analysis and marketing actions.

Research Problem Solution

The authors proposed a method for market segmentation that is performed as per the customer lifestyle, including the information attained from an extensive transactional-based database in the retail domain. First, the data is obtained using a variable clustering algorithm to mine data from the shopping baskets, which are further used to understand the customer lifestyles. The identified information from shopping baskets includes products that are frequently purchased together. Furthermore, the customers’ lifestyle can be assumed depending on the different products purchased over a period, and customers are assigned to various lifestyle-based segments based on the history of their purchases.

Cluster analysis was implemented to map data items into groups with high similarity (i.e., clusters). However, the author suggested that lifestyle profiles are more stable compared to customer groups; therefore, the authors implemented a variable clustering algorithm instead of a general clustering algorithm used mainly for customer group-based segmentation tasks. A VARCLUS algorithm was integrated with the help of SAS software for clustering according to the products. The VARCLUS algorithm is based on the divisive algorithm wherein the items are divided into two groups in its first iteration. In addition, the successive iterations include all the remaining items of each group in the cluster. However, in this process, the groups are split only when there is a specific variable to be explained with the help of splitting. On the other hand, each item is provided with a unique cluster, with a possibility to reassign the item to another cluster if required in the subsequent iteration, unlike the approach in a standard divisive algorithm.

On obtaining the product-based clusters using the VACLUS algorithm, the product features of each cluster are analyzed to identify the lifestyle of the customers buying the products. To ensure generality of the outcome, the analysis focused only on the business units, the categories, the product position of the concerned brand, and the value after identifying the lifestyle segments. Upon identifying the lifestyle-based segments, the customers are assigned to these segments based on the shopping basket information where the products are similar to the corresponding past purchases.

As for the lifestyle segments, the analysis reported by the study is based on the transactional-based data obtained from customers owning loyalty cards. This database includes records from October to November 2009. The lifestyle segments were performed using an exploratory analysis that included the data from different sources while eliminating any outliers in the data. There was a total of ten products selected from the two months analyzed. In addition, the resulting sample included 100,000 randomly selected customers for the variable clustering procedure. However, the authors suggested that their primary interest was to analyze representative products related to sales.

Further, from the 105,160 products purchased, only those products were selected that were purchased by at least 10,000 customers. Based on the transaction records, the binary matrix comprising items in the shopping baskets was created for 100,000 customers and 1831 products. The binary structure provided insight into whether the customer did or did not buy each product. The primary motive of the analysis centered around the investigation of the relationship between different products, regardless of the quantities. Therefore, no matrix was created relating to the quantities bought by each customer and each product. Figure.1 provides the information on the possible and appropriate number of clusters created to ensure that the segmentation could provide vital information for marketing actions.

Figure.1 Products in Each Cluster

The authors indicated that the criteria are essential for the segments to be influential. Hence, the products were grouped in six clusters. From the 1831 products, each product is assigned to one of the six clusters. In order to identify the purchase patterns of the customers related to the cluster of products, the products’ business units were considered that were present in each cluster. For that purpose, the computational analysis included the ratio between the proportion of products in a specific cluster that belonged to each business unit, including the average proportion of the products of that particular business unit in the sample used for analysis. The results obtained from the analysis are highlighted in Figure 2.

Figure.2 Products Available in Each Business Unit

Findings of the Research

The study sheds light on the identified lifestyle of the buyers of each group of products (see Figure 3).

Figure.3 Products in Each Cluster

Cluster 1 represents a relatively higher number of bakery items to takeaways in the above figure. However, this cluster is significantly related to the customer with medium purchasing power and mainly focuses on purchases of practical meal requirements. On the other hand, these buyers prefer takeaways and products for preparing food items like sandwiches. Besides, these customers have babies and have a liking for wine.

Cluster 2 is associated with fruits, vegetables, bakery products, cookies, fresh fish items, and wine or champagne. However, this cluster is also related to secondary or own-brand products. In addition, the buyers’ purchasing power in this cluster is of medium category, while they have a preference for a balanced diet and socializing identified from the diverse range of fortified wine and champagnes purchased in a given period.

At the same time, cluster 3 represents purchases of products like drinks, fruits, vegetables, grocery, and butchery business items. This cluster’s most relevant products include appetizers, spirit-based drinks, veal meat, liquid fats, perfumes, cosmetics, dishwashing products, cheese on the counter, fruits, and frozen food vegetable items. Similarly, the products are also from a secondary brand or own brand. However, the buyers’ purchasing power is in the medium category with a high preference for meat products, care for personal appearance, and enjoy socializing.

Cluster 4 includes buyers with a preference for products such as hygiene, grocery, butchery products, meat on the counter, canned food, pet care, and basic amenities such as honey, jam, and eggs. However, most products are of a premium brand or an economical brand. On the contrary, the author suggests that the buyers of this group have low purchasing power even if the buyers prefer premium brands. In addition, the customers have likeliness for pets and often prepare dishes that require essential ingredients.

Cluster 5 includes drinks, hygiene, dairy, frozen fishery, and butchery items. The relevant categories of products include health care, frozen fish, baby hygiene, oral hygiene products, perfumes, consumables, men’s products, barbecue chicken, pork meat, spice, and wines. This cluster purchase from premium brands or leader brands and have high purchasing power. Therefore, the customers in this group have babies, are financially well-equipped, prefer frozen products and barbecue and wine, and are partially invested in health, hygiene, and cosmetic necessities.

The final cluster 6 includes baby products, codfish, powdered brinks, cereals, fats, frozen meals, and pre-cooked items. The buyers of this group buy from premium brands or their own brands. These customers of this cluster have high economic power and have babies and prefer practical meal solutions and cod fish-based meals. However, the economic power of the buyers in the group is lower when compared to buyers of cluster 5 (see Figure 4). Based on the outcome of the analysis, the authors suggest that the study’s lifestyle segmentation can contribute to designing company strategies and actions.

Figure.4 Product Proportions in the Primary Category

Future Work Suggestions and Implications for Practitioners

The authors highlighted that a well-defined market segmentation ensures companies build a solid relationship with their customers leading to higher sales and profits. Therefore, the study provided a direction for managerial policies for practitioners that can be implemented based on the outcome of the analytical results of the lifestyle segmentation. Furthermore, the results signify that there is a possibility of using segmentation for promotional campaigns because lifestyle segmentation provides a more straightforward identification of customers’ interest in a particular product. Therefore, there are higher possibilities for the campaign to be more successful if understanding the customer needs and products is offered by understanding their interests.

As a result, the range of products in the store can be managed by considering the segments representative of each store. Moreover, it allows stores to include a diverse range of products belonging to each category of customer segments and their preferences, which leads to increased sales and higher customer satisfaction. In addition, the store layout can be well-defined with more categories for the client segments that visit the stores often in different regions. This scenario enables stores to increase sales by offering more convenience to customers. Finally, the authors suggest that customer segmentation is key to gaining a competitive advantage and enables companies to create well-defined strategic actions targeted at the benefit of the customers and the company. Consequently, recognizing the customer differences ensures successful marketing campaigns, with higher satisfaction of customer needs that can build loyal relationships with a wide range of customer base.