MediAnalytics: Integration Practice of Data Analysis and Machine Learning in Pharmaceutical Retail

MediAnalytics is a data analysis solution for B2C pharmaceutical retail scenarios, integrating Power BI visualization and Python machine learning technologies to achieve core functions such as sales insights, customer churn prediction, and delivery expansion analysis. Built on one year of operational data from a real pharmaceutical retail store, the project demonstrates the practical value of data-driven thinking in vertical industries, providing references for enterprise decision-making, analyst learning, and technical practice.

The pharmaceutical retail industry is undergoing digital transformation, and traditional experience-driven business models struggle to meet complex market demands. B2C pharmaceutical retail enterprises face challenges such as multi-source data integration, business insight mining, and customer behavior prediction. The MediAnalytics project addresses these pain points by building a complete data analysis and machine learning solution based on one year of operational data from a real single store.

The project integrates multi-source heterogeneous data:

1. Purchase record data: includes basic dimensions for sales analysis such as transaction time, product information, and amount;
2. User profile data: includes demographics, registration time, geographic location, etc., as the basis for customer analysis;
3. Drug information data: includes classification, efficacy, inventory status, etc., to support correlation analysis. For privacy and security reasons, only a small amount of sample data is displayed, and the complete dataset is not publicly available.

The project builds six core modules:

1. Sales analysis dashboard: multi-dimensional (time/product/region/crowd) sales trend and feature analysis;
2. Geospatial insight: geographic distribution visualization to support store location selection and delivery optimization;
3. Pareto analysis: apply the 80/20 rule to identify core products and optimize inventory allocation;
4. Customer churn prediction: use logistic regression, random forest, and XGBoost models to intervene in high-risk customers in advance;
5. Delivery expansion analysis: evaluate the balance between demand and cost in new regions to support delivery scope decisions;
6. Discount effect testing: optimize promotion strategies through statistical tests.

A hybrid technology stack is used to implement the entire process:

• Data processing layer: Python (Pandas for cleaning/transformation, NumPy for calculation, Matplotlib for visualization);
• Visualization layer: Power BI (DAX indicator calculation, slicer dynamic filtering, map geographic display);
• Machine learning layer: Scikit-Learn (classification models, statistical tests);
• Data storage: Excel (adapted to the lightweight needs of small and medium-sized enterprises).

• Pharmaceutical retail enterprises: Establish data-driven decision-making mechanisms, optimize inventory/purchasing/customer retention, and support delivery planning;
• Data analysts: Learn vertical industry analysis ideas and master Power BI and Python integration methods;
• Technical learners: Understand business-to-technology solutions and gain end-to-end project practical experience.

Limitations: Limited data scale (one year of a single store), lack of real-time data flow, insufficient model depth, weak interpretability; Improvement directions: Upgrade data infrastructure to databases/data warehouses, integrate stream processing to achieve real-time analysis, introduce time series prediction/recommendation systems, establish MLOps processes, and integrate SHAP/LIME to improve model transparency.

MediAnalytics demonstrates the value of combining general technologies with vertical business scenarios, and its methodology can be migrated to other retail formats. In the pharmaceutical retail industry, data capability is becoming a core competitiveness. In the future, AI technology will promote more intelligent solutions, making data an engine for business growth.