The COVID-19 pandemic instigated thousands of companies' closures and affected offline retail shops. Thus, online B2C business models enable traditional offline stores to boost their sales. This study aims to explore the use of historical sales and behavioral data analytics to construct a recommendation model. A process model is proposed, which is the combination of recency, frequency, and monetary (RFM) analysis method and the k-means clustering algorithm. RFM analysis is used to segment customer levels in the company while the association rule theory and the apriori algorithm are utilized for completing the shopping basket analysis and recommending products based on the results. The proposed recommendation model provides a good marketing mix to improve sales and market responsiveness. In addition, it recommends specific products to new customers as well as specific groups of target customers. This study offered a practical business transformation case that can assist companies in a similar situation to transform their business model and improve their profits.
TopIntroduction
Advances in computing have allowed faster deployment and better optimization of machine learning (ML), which has the potential to be applied in all types of industries today (Abdelrahman & Keikhosrokiani, 2020; Keikhosrokiani, 2022; Keikhosrokiani & Asl, 2022; Teoh Yi Zhe & Keikhosrokiani, 2021). In the e-business world, online shopping has already become a popular shopping trading methods in more and more countries. The advanced algorithms used in recommender system and customer value management will make more customer-oriented marketing strategy, which improve customer satisfaction and corporate profits at the same time. Improved recommendation and prediction models, which are based on user behavior analysis (UBA), have greatly benefited the retail industry. However, many traditional offline stores don’t have time or chance to transform their business model to online stores or combination model due to the sudden COVID-19 pandemic.
Data mining is the most effective way to analyze customer purchase behavior, and it can discover hidden useful information from massive online transaction data (Ngai et al., 2009; Pantea Keikhosrokiani, 2021; Shaw et al., 2001). A start-up e-commerce company can utilize data mining techniques to find customers’ favorites commodities, such as books, music records, clothes, electronic equipment or foods from the historical shopping data. In addition, this is also conducive to the company’s optimization inventory to avoid the backlog of goods. Conversely, hot goods can also be explicitly increased to maximize sales, avoid insufficient inventory, thereby reducing sales stagnation due to outstanding stocks. Therefore, a start-up cosmetics e-commerce company must establish and implement a system to predict sales and product recommendation. Conversely, it is also possible to add popular items or similar items to maximize sales and avoid running out of stock, thus reducing sales stagnation due to product out-of-stock (Sarwar et al., 2000). Therefore, a start-up cosmetics e-commerce company must implement a system to forecast sales and product recommendations.
Traditional offline shops usually face three problems: (1) limited market, (2) long sale cycle, and (3) complex and time-consuming sale process due to the current market situation affected by COVID-19 pandemic. In addition, they generally face two types of customers. One is general public consumer that their purchase decision usually involves two or three decision makers such as themselves, their family, or their friends. As a result, the total time for a purchase decision is usually very short. Another is the big customer in which their B2B sales cycles involve a complex set of factors involving multiple stakeholders, purchasing process decision makers, professional purchasing teams, experts from different positions or fields who have more specialized and process-oriented operations. For the big customer group, many business negotiations would be carried out and therefore, it is not surprising that decision times are usually longer up to several months. Longer sales cycles pose a company’s cash flow which establishes a huge challenge that many companies must face, but maybe it is critical for a company seeking to transformation.
In compared to the e-commerce industry, more traditional industries are gradually paying attention to the importance of customer segmentation. For example, in the telecommunications industry, which is a data-intensive field, communication customers will continue to generate a lot of data. Therefore, customer segmentation is one of the most important applications in data mining in the telecommunications industry. Customer segmentation methods divides customers into different groups according to customer properties. The customer’s similarities with more similarities will be segmented in the same group whereas the customers in different groups have the many difference (Qiu et al., 2020). Then, based on customers’ characteristics, it is more likely to explore whether elder customers prefer to call, young customers prefer data traffic to watch videos, play games, etc.