Data Analytics Predictive Model to Promote Up-Selling and Cross-Selling Activities Based on Consumer’s Buying Pattern

Abstract

In the current competitive market, companies need to utilise data-driven approaches to maximise sales and improve customer interaction. This research intends to create a predictive analytics and recommendation system to enhance up-selling and cross-selling efforts for dealers. By employing machine learning models such as linear regression, random forest, and extreme gradient boosting (XGBoost), the system predicts product demand, facilitating improved inventory control and sales enhancement. Moreover, an association rule mining (Apriori algorithm) method is utilised to uncover frequent item associations, producing customised product suggestions to increase transaction values. Research results showed that the most suitable prediction method is XGBoost for this research case, with an accuracy of R² = 0.810543. The research solution is proven to be scalable through data analytics using three different synthetic datasets from Kaggle, including Rossmann Store Sales, Bakery Transaction Data, and Corporación Favorita Store Sales. The research integrates these elements into an interactive Power BI dashboard, providing dealers instant insights into customer purchase patterns, sales outcomes, and inventory trends. This research provides a one-stop solution for dealers to observe sales trends, and with insights into recommended items, dealers can achieve up-selling and cross-selling. Key performance indicators (KPI) are shown in the Power BI dashboard as a centralised display, thus dealers can make decisions to promote products to customers with a higher success rate. In future, the research will hopefully expand data sources to industrial data, improve model performance, and integrate more advanced technology into the Dealer Management Systems (DMS). The DMS developed is scalable and can deal with expansion and upscale upgrades in future work.